update get_resized_cond to match new cond format

update sampling function to match new ComfyUI update
fix: new comfyui break sampling
2023-10-25 22:11:13 +07:00 · 2023-10-25 16:30:06 +07:00 · 2023-10-25 15:24:50 +07:00 · 2023-10-12 21:46:57 +07:00 · 2023-10-12 21:44:21 +07:00 · 2023-09-25 22:59:21 +07:00
16 changed files with 2712 additions and 2797 deletions
@@ -11,6 +11,54 @@
 - Community modules: [manshoety/AD_Stabilized_Motion](https://huggingface.co/manshoety/AD_Stabilized_Motion) | [CiaraRowles/TemporalDiff](https://huggingface.co/CiaraRowles/TemporalDiff)
 - AnimateDiff v2 [mm_sd_v15_v2.ckpt](https://huggingface.co/guoyww/animatediff/blob/main/mm_sd_v15_v2.ckpt)

+## Update 2023/09/25
+
+#### **Motion LoRA** is now supported!
+
+Download [motion LoRAs](https://huggingface.co/guoyww/animatediff/tree/main) and put them under `comfyui-animatediff/loras/` folder.
+
+Note: LoRAs only work with **AnimateDiff v2** [mm_sd_v15_v2.ckpt](https://huggingface.co/guoyww/animatediff/blob/main/mm_sd_v15_v2.ckpt) module.
+
+#### New node: `AnimateDiffLoraLoader`
+
+<img width="370" alt="image" src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/7a9f62f7-702e-48a4-934c-bbfe1e23aff2">
+
+Example workflow:
+<img width="1280" alt="image" src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/93e7550f-4648-4482-9961-6cece5132dc9">
+
+Workflow: [lora.json](https://github.com/ArtVentureX/comfyui-animatediff/blob/main/workflows/lora.json)
+
+Samples:
+
+<table>
+<tr>
+<td>
+<img width="512" alt="image" src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/2c5aa25e-0682-481f-8842-066c5b988864">
+</td>
+</tr>
+<tr>
+<td>
+<img width="512" alt="image" src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/adfbad45-3ba5-42e3-9bee-d2b83f43989c">
+</td>
+</tr>
+<tr>
+<td>
+<img width="512" alt="image" src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/8e484c74-c691-4d1c-9514-719dbfe3a0b5">
+</td>
+</tr>
+<tr>
+<td>
+<img width="512" alt="image" src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/4921a335-9207-4a7b-9d66-61a5d76e3179">
+</td>
+</tr>
+</table>
+
+## Update 2023/09/21
+
+#### **Sliding Window** is now available!
+
+The sliding window feature enables you to generate GIFs without a frame length limit. It divides frames into smaller batches with a slight overlap. This feature is activated automatically when generating more than 16 frames. To modify the trigger number and other settings, utilize the `SlidingWindowOptions` node. See the [sample workflow](#long-duration-with-sliding-window) bellow.
+
 ## Nodes

 #### AnimateDiffLoader
@@ -20,12 +68,13 @@
 #### AnimateDiffSampler

 - Mostly the same with `KSampler`
- Use `AnimateDiffLoader` to load the motion module
+- `motion_module`: use `AnimateDiffLoader` to load the motion module
 - `inject_method`: should left default
 - `frame_number`: animation length
 - `latent_image`: You can pass an `EmptyLatentImage`
+- `sliding_window_opts`: custom sliding window options

-<img width="370" alt="image" src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/f22d6b36-ce36-44cc-80e8-dffe6f77b296">
+<img width="370" alt="image" src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/a352195d-f40c-494d-bd3d-30ee88174b88">

 #### AnimateDiffCombine

@@ -33,10 +82,34 @@
 - `frame_rate`: number of frame per second
 - `loop_count`: use 0 for infinite loop
 - `save_image`: should GIF be saved to disk
- `format`: supports `image/gif`, `image/webp` (better compression) or `video/webm` (need `ffmpeg` installed and available in PATH)
+- `format`: supports `image/gif`, `image/webp` (better compression), `video/webm`, `video/h264-mp4`, `video/h265-mp4`. To use video formats, you'll need [ffmpeg](https://ffmpeg.org/download.html) installed and available in **`PATH`**

 <img width="370" alt="image" src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/381c5acc-06ef-43da-ada0-3dc76f37a3e4">

+#### SlidingWindowOptions
+
+Custom sliding window options
+
+- `context_length`: number of frame per _window_. Use **16** to get the best results. Reduce it if you have low VRAM.
+- `context_stride`:
+  - 1: sampling every frame
+  - 2: sampling every frame then every second frame
+  - 3: sampling every frame then every second frame then every third frames
+  - ...
+- `context_overlap`: overlap frames between each window slice
+- `closed_loop`: make the GIF a closed loop, will add more sampling step
+
+<img width="370" alt="image" src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/6679a8dd-bf96-419f-8934-ea2b046dd23c">
+
+#### LoadVideo
+
+Load GIF or video as images. Usefull to load a GIF as ControlNet input.
+
+- `frame_start`: Skip some begining frames and start at `frame_start`
+- `frame_limit`: Only take `frame_limit` frames
+
+<img width="370" alt="image" src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/684176d5-6369-4a27-9f33-e721e0fe1876">
+
 ## Workflows

 ### Simple txt2gif
@@ -51,6 +124,27 @@ Samples:

 ![animate_diff_02](https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/c39b26f7-a2af-4dc4-902f-c363e2e6f39a)

+### Long duration with sliding window
+
+<img width="1280" alt="image" src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/0f8bfb87-83cb-4119-9777-e3948ec0cb5c">
+
+Workflow: [sliding-window.json](https://github.com/ArtVentureX/comfyui-animatediff/blob/main/workflows/sliding-window.json)
+
+Samples:
+
+<table>
+<tr>
+<td>
+<img width="512" alt="image" src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/e1da7a66-e615-475d-9400-41eff484ad49">
+</td>
+</tr>
+<tr>
+<td>
+<img width="768" alt="image" src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/4faa7e5e-cdaa-49da-8759-46d779c0e0b6">
+</td>
+</tr>
+</table>
+
 ### Latent upscale

 Upscale latent output using `LatentUpscale` then do a 2nd pass with `AnimateDiffSampler`.
@@ -122,6 +216,10 @@ Samples:

 ## Known Issues

+### CUDA error: invalid configuration argument
+
+It's an `xformers` bug accidentally triggered by the way the original AnimateDiff CrossAttention is passed in. The current workaround is to disable xformers with `--disable-xformers` when booting ComfyUI.
+
 ### GIF split into multiple scenes

 ![AnimateDiff_00007_](https://github.com/ArtVentureX/comfyui-animatediff/assets/8894763/e6cd53cb-9878-45da-a58a-a15851882386)
@@ -1,7 +1,18 @@
 import os
 import hashlib
+import torch
+from typing import Dict

 import folder_paths
+import comfy.model_management as model_management
+from comfy.utils import load_torch_file, calculate_parameters
+
+from .logger import logger
+from .motion_module import MotionWrapper
+
+
+motion_modules: Dict[str, MotionWrapper] = {}
+motion_loras: Dict[str, Dict[str, torch.Tensor]] = {}


 folder_paths.folder_names_and_paths["AnimateDiff"] = (
@@ -11,11 +22,12 @@ folder_paths.folder_names_and_paths["AnimateDiff"] = (
    ],
    folder_paths.supported_pt_extensions,
 )
-folder_paths.folder_names_and_paths["video_formats"] = (
+folder_paths.folder_names_and_paths["AnimateDiffLora"] = (
    [
-        os.path.join(os.path.dirname(os.path.abspath(__file__)), "..", "video_formats"),
+        os.path.join(folder_paths.models_dir, "AnimateDiffLora"),
+        os.path.join(os.path.dirname(os.path.abspath(__file__)), "..", "loras"),
    ],
-    [".json"]
+    folder_paths.supported_pt_extensions,
 )


@@ -23,11 +35,66 @@ def get_available_models():
    return folder_paths.get_filename_list("AnimateDiff")


+def get_available_loras():
+    return folder_paths.get_filename_list("AnimateDiffLora")
+
+
 def get_model_path(model_name):
    return folder_paths.get_full_path("AnimateDiff", model_name)


+def get_lora_path(lora_name):
+    return folder_paths.get_full_path("AnimateDiffLora", lora_name)
+
+
 def get_model_hash(file_path):
    with open(file_path, "rb") as f:
-        bytes = f.read()  # read entire file as bytes
+        bytes = f.read(1024 * 1024)  # read entire file as bytes
        return hashlib.sha256(bytes).hexdigest()
+
+
+def load_motion_module(model_name: str):
+    model_path = get_model_path(model_name)
+    model_hash = get_model_hash(model_path)
+    if model_hash not in motion_modules:
+        logger.info(f"Loading motion module {model_name}")
+        mm_state_dict = load_torch_file(model_path)
+        motion_module = MotionWrapper.from_state_dict(mm_state_dict, model_name)
+
+        params = calculate_parameters(mm_state_dict, "")
+        if model_management.should_use_fp16(model_params=params):
+            logger.info(f"Converting motion module to fp16.")
+            motion_module.half()
+        offload_device = model_management.unet_offload_device()
+        motion_module = motion_module.to(offload_device)
+
+        motion_modules[model_hash] = motion_module
+
+    return motion_modules[model_hash]
+
+
+def load_lora(lora_name: str):
+    lora_path = get_lora_path(lora_name)
+    lora_hash = get_model_hash(lora_path)
+    if lora_hash not in motion_modules:
+        logger.info(f"Loading lora {lora_name}")
+        state_dict = load_torch_file(lora_path)
+        updated_state_dict: Dict[str, torch.Tensor] = {}
+
+        for key in state_dict:
+            # only process lora down key
+            if "up." in key:
+                continue
+
+            up_key = key.replace(".down.", ".up.")
+            model_key = key.replace("processor.", "").replace("_lora", "").replace("down.", "").replace("up.", "")
+            model_key = model_key.replace("to_out.", "to_out.0.")
+            combined_key = ".".join(model_key.split(".")[:-1])
+
+            weight_down = state_dict[key]
+            weight_up = state_dict[up_key]
+            updated_state_dict[combined_key] = torch.mm(weight_up, weight_down).to("cpu")
+
+        motion_loras[lora_hash] = updated_state_dict
+
+    return motion_loras[lora_hash]
@@ -1,11 +1,35 @@
-import os
 import torch
 from torch import Tensor, nn

 import math
 from einops import rearrange, repeat

-from comfy.ldm.modules.attention import FeedForward, CrossAttention
+import comfy.model_management as model_management
+from comfy.ldm.modules.attention import (
+    default,
+    FeedForward,
+    CrossAttention as ComfyCrossAttention,
+    attention_basic,
+    attention_pytorch,
+    attention_split,
+    attention_sub_quad,
+)
+from comfy.cli_args import args
+
+from .logger import logger
+
+attention = attention_basic
+
+if model_management.xformers_enabled():
+    logger.warn("xformers is enabled but it has a bug that can cause issue while using with AnimateDiff.")
+
+if model_management.pytorch_attention_enabled():
+    attention = attention_pytorch
+else:
+    if args.use_split_cross_attention:
+        attention = attention_split
+    else:
+        attention = attention_sub_quad


 def zero_module(module):
@@ -32,6 +56,24 @@ def has_mid_block(mm_state_dict: dict[str, Tensor]):
    return False


+class CrossAttention(ComfyCrossAttention):
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+
+    def forward(self, x, context=None, value=None, mask=None):
+        q = self.to_q(x)
+        context = default(context, x)
+        k = self.to_k(context)
+        if value is not None:
+            v = self.to_v(value)
+            del value
+        else:
+            v = self.to_v(context)
+
+        out = attention(q, k, v, self.heads, mask)
+        return self.to_out(out)
+
+
 class MotionWrapper(nn.Module):
    def __init__(self, mm_type: str, encoding_max_len: int = 24, is_v2=False):
        super().__init__()
@@ -41,22 +83,17 @@ class MotionWrapper(nn.Module):
        self.down_blocks = nn.ModuleList([])
        self.up_blocks = nn.ModuleList([])
        self.mid_block = None
+        self.encoding_max_len = encoding_max_len

        for c in (320, 640, 1280, 1280):
-            self.down_blocks.append(
-                MotionModule(c, BlockType.DOWN, encoding_max_len=encoding_max_len)
-            )
+            self.down_blocks.append(MotionModule(c, BlockType.DOWN, encoding_max_len=encoding_max_len))
        for c in (1280, 1280, 640, 320):
-            self.up_blocks.append(
-                MotionModule(c, BlockType.UP, encoding_max_len=encoding_max_len)
-            )
+            self.up_blocks.append(MotionModule(c, BlockType.UP, encoding_max_len=encoding_max_len))
        if is_v2:
-            self.mid_block = MotionModule(
-                1280, BlockType.MID, encoding_max_len=encoding_max_len
-            )
+            self.mid_block = MotionModule(1280, BlockType.MID, encoding_max_len=encoding_max_len)

    @classmethod
-    def from_pretrained(cls, mm_state_dict: dict[str, Tensor], mm_type: str):
+    def from_state_dict(cls, mm_state_dict: dict[str, Tensor], mm_type: str):
        encoding_max_len = get_encoding_max_len(mm_state_dict)
        is_v2 = has_mid_block(mm_state_dict)

@@ -90,9 +127,7 @@ class MotionModule(nn.Module):
        self.block_type = block_type

        if block_type == BlockType.MID:
-            self.motion_modules = nn.ModuleList(
-                [get_motion_module(in_channels, encoding_max_len)]
-            )
+            self.motion_modules = nn.ModuleList([get_motion_module(in_channels, encoding_max_len)])
        else:
            self.motion_modules = nn.ModuleList(
                [
@@ -101,9 +136,7 @@ class MotionModule(nn.Module):
                ]
            )
            if block_type == BlockType.UP:
-                self.motion_modules.append(
-                    get_motion_module(in_channels, encoding_max_len)
-                )
+                self.motion_modules.append(get_motion_module(in_channels, encoding_max_len))

    def set_video_length(self, video_length: int):
        for motion_module in self.motion_modules:
@@ -111,9 +144,7 @@ class MotionModule(nn.Module):


 def get_motion_module(in_channels, max_len):
-    return VanillaTemporalModule(
-        in_channels=in_channels, temporal_position_encoding_max_len=max_len
-    )
+    return VanillaTemporalModule(in_channels=in_channels, temporal_position_encoding_max_len=max_len)


 class VanillaTemporalModule(nn.Module):
@@ -134,9 +165,7 @@ class VanillaTemporalModule(nn.Module):
        self.temporal_transformer = TemporalTransformer3DModel(
            in_channels=in_channels,
            num_attention_heads=num_attention_heads,
-            attention_head_dim=in_channels
-            // num_attention_heads
-            // temporal_attention_dim_div,
+            attention_head_dim=in_channels // num_attention_heads // temporal_attention_dim_div,
            num_layers=num_transformer_block,
            attention_block_types=attention_block_types,
            cross_frame_attention_mode=cross_frame_attention_mode,
@@ -145,17 +174,13 @@ class VanillaTemporalModule(nn.Module):
        )

        if zero_initialize:
-            self.temporal_transformer.proj_out = zero_module(
-                self.temporal_transformer.proj_out
-            )
+            self.temporal_transformer.proj_out = zero_module(self.temporal_transformer.proj_out)

    def set_video_length(self, video_length: int):
        self.temporal_transformer.set_video_length(video_length)

-    def forward(self, input_tensor, encoder_hidden_states, attention_mask=None):
-        return self.temporal_transformer(
-            input_tensor, encoder_hidden_states, attention_mask
-        )
+    def forward(self, input_tensor, encoder_hidden_states=None, attention_mask=None):
+        return self.temporal_transformer(input_tensor, encoder_hidden_states, attention_mask)


 class TemporalTransformer3DModel(nn.Module):
@@ -183,9 +208,7 @@ class TemporalTransformer3DModel(nn.Module):

        inner_dim = num_attention_heads * attention_head_dim

-        self.norm = torch.nn.GroupNorm(
-            num_groups=norm_num_groups, num_channels=in_channels, eps=1e-6, affine=True
-        )
+        self.norm = torch.nn.GroupNorm(num_groups=norm_num_groups, num_channels=in_channels, eps=1e-6, affine=True)
        self.proj_in = nn.Linear(in_channels, inner_dim)

        self.transformer_blocks = nn.ModuleList(
@@ -220,9 +243,7 @@ class TemporalTransformer3DModel(nn.Module):

        hidden_states = self.norm(hidden_states)
        inner_dim = hidden_states.shape[1]
-        hidden_states = hidden_states.permute(0, 2, 3, 1).reshape(
-            batch, height * weight, inner_dim
-        )
+        hidden_states = hidden_states.permute(0, 2, 3, 1).reshape(batch, height * weight, inner_dim)
        hidden_states = self.proj_in(hidden_states)

        # Transformer Blocks
@@ -235,11 +256,7 @@ class TemporalTransformer3DModel(nn.Module):

        # output
        hidden_states = self.proj_out(hidden_states)
-        hidden_states = (
-            hidden_states.reshape(batch, height, weight, inner_dim)
-            .permute(0, 3, 1, 2)
-            .contiguous()
-        )
+        hidden_states = hidden_states.reshape(batch, height, weight, inner_dim).permute(0, 3, 1, 2).contiguous()

        output = hidden_states + residual

@@ -275,9 +292,7 @@ class TemporalTransformerBlock(nn.Module):
            attention_blocks.append(
                VersatileAttention(
                    attention_mode=block_name.split("_")[0],
-                    context_dim=cross_attention_dim
-                    if block_name.endswith("_Cross")
-                    else None,
+                    context_dim=cross_attention_dim if block_name.endswith("_Cross") else None,
                    query_dim=dim,
                    heads=num_attention_heads,
                    dim_head=attention_head_dim,
@@ -309,9 +324,7 @@ class TemporalTransformerBlock(nn.Module):
            hidden_states = (
                attention_block(
                    norm_hidden_states,
-                    encoder_hidden_states=encoder_hidden_states
-                    if attention_block.is_cross_attention
-                    else None,
+                    encoder_hidden_states=encoder_hidden_states if attention_block.is_cross_attention else None,
                    video_length=video_length,
                )
                + hidden_states
@@ -328,9 +341,7 @@ class PositionalEncoding(nn.Module):
        super().__init__()
        self.dropout = nn.Dropout(p=dropout)
        position = torch.arange(max_len).unsqueeze(1)
-        div_term = torch.exp(
-            torch.arange(0, d_model, 2) * (-math.log(10000.0) / d_model)
-        )
+        div_term = torch.exp(torch.arange(0, d_model, 2) * (-math.log(10000.0) / d_model))
        pe = torch.zeros(1, max_len, d_model)
        pe[0, :, 0::2] = torch.sin(position * div_term)
        pe[0, :, 1::2] = torch.cos(position * div_term)
@@ -382,9 +393,7 @@ class VersatileAttention(CrossAttention):
            raise NotImplementedError

        d = hidden_states.shape[1]
-        hidden_states = rearrange(
-            hidden_states, "(b f) d c -> (b d) f c", f=video_length
-        )
+        hidden_states = rearrange(hidden_states, "(b f) d c -> (b d) f c", f=video_length)

        if self.pos_encoder is not None:
            hidden_states = self.pos_encoder(hidden_states)
@@ -3,176 +3,24 @@ import json
 import torch
 import numpy as np
 import hashlib
-from typing import Dict, List
+from typing import List, Dict, Tuple
 from torch import Tensor
-from torch.nn.functional import group_norm
 from PIL import Image, ImageSequence
 from PIL.PngImagePlugin import PngInfo
-from einops import rearrange

 import folder_paths
-import comfy.ldm.modules.diffusionmodules.openaimodel as openaimodel
-import comfy.model_management as model_management
-from comfy.model_base import BaseModel
-from comfy.ldm.modules.attention import SpatialTransformer
-from comfy.utils import load_torch_file, calculate_parameters
-from nodes import KSampler

+from .motion_module import MotionWrapper
+from .model_utils import get_available_models, load_motion_module, get_available_loras, load_lora
+from .utils import pil2tensor, ensure_opencv
+from .sampler import AnimateDiffSampler, AnimateDiffSlidingWindowOptions
 from .logger import logger
-from .motion_module import MotionWrapper, VanillaTemporalModule
-from .model_utils import get_available_models, get_model_path, get_model_hash
-from .utils import pil2tensor


-def forward_timestep_embed(
-    ts, x, emb, context=None, transformer_options={}, output_shape=None
-):
-    for layer in ts:
-        if isinstance(layer, openaimodel.TimestepBlock):
-            x = layer(x, emb)
-        elif isinstance(layer, VanillaTemporalModule):
-            x = layer(x, context)
-        elif isinstance(layer, SpatialTransformer):
-            x = layer(x, context, transformer_options)
-            transformer_options["current_index"] += 1
-        elif isinstance(layer, openaimodel.Upsample):
-            x = layer(x, output_shape=output_shape)
-        else:
-            x = layer(x)
-    return x
+SLIDING_CONTEXT_LENGTH = 16

-
-def groupnorm_mm_factory(video_length: int):
-    def groupnorm_mm_forward(self, input: Tensor) -> Tensor:
-        # axes_factor normalizes batch based on total conds and unconds passed in batch;
-        # the conds and unconds per batch can change based on VRAM optimizations that may kick in
-        axes_factor = input.size(0) // video_length
-
-        input = rearrange(input, "(b f) c h w -> b c f h w", b=axes_factor)
-        input = group_norm(input, self.num_groups,
-                           self.weight, self.bias, self.eps)
-        input = rearrange(input, "b c f h w -> (b f) c h w", b=axes_factor)
-        return input
-
-    return groupnorm_mm_forward
-
-
-orig_forward_timestep_embed = openaimodel.forward_timestep_embed
-orig_maximum_batch_area = model_management.maximum_batch_area
-orig_groupnorm_forward = torch.nn.GroupNorm.forward
-openaimodel.forward_timestep_embed = forward_timestep_embed
-
-motion_modules: Dict[str, MotionWrapper] = {}
-
-
-def load_motion_module(model_name: str):
-    model_path = get_model_path(model_name)
-    model_hash = get_model_hash(model_path)
-    if model_hash not in motion_modules:
-        logger.info(f"Loading motion module {model_name}")
-        mm_state_dict = load_torch_file(model_path)
-        motion_module = MotionWrapper.from_pretrained(
-            mm_state_dict, model_name)
-
-        params = calculate_parameters(mm_state_dict, "")
-        if model_management.should_use_fp16(model_params=params):
-            logger.info(f"Converting motion module to fp16.")
-            motion_module.half()
-        offload_device = model_management.unet_offload_device()
-        motion_module = motion_module.to(offload_device)
-
-        motion_modules[model_hash] = motion_module
-
-    return motion_modules[model_hash]
-
-
-def inject_motion_module_to_unet_legacy(unet, motion_module: MotionWrapper):
-    for mm_idx, unet_idx in enumerate([1, 2, 4, 5, 7, 8, 10, 11]):
-        mm_idx0, mm_idx1 = mm_idx // 2, mm_idx % 2
-        unet.input_blocks[unet_idx].append(
-            motion_module.down_blocks[mm_idx0].motion_modules[mm_idx1]
-        )
-
-    for unet_idx in range(12):
-        mm_idx0, mm_idx1 = unet_idx // 3, unet_idx % 3
-        if unet_idx % 2 == 2:
-            unet.output_blocks[unet_idx].insert(
-                -1, motion_module.up_blocks[mm_idx0].motion_modules[mm_idx1]
-            )
-        else:
-            unet.output_blocks[unet_idx].append(
-                motion_module.up_blocks[mm_idx0].motion_modules[mm_idx1]
-            )
-    if motion_module.is_v2:
-        unet.middle_block.insert(-1, motion_module.mid_block.motion_modules[0])
-
-    unet.motion_module = motion_module
-
-
-def eject_motion_module_from_unet_legacy(unet):
-    for unet_idx in [1, 2, 4, 5, 7, 8, 10, 11]:
-        unet.input_blocks[unet_idx].pop(-1)
-
-    for unet_idx in range(12):
-        if unet_idx % 2 == 2:
-            unet.output_blocks[unet_idx].pop(-2)
-        else:
-            unet.output_blocks[unet_idx].pop(-1)
-
-    if unet.motion_module.is_v2:
-        unet.middle_block.pop(-2)
-
-    del unet.motion_module
-
-
-def inject_motion_module_to_unet(unet, motion_module: MotionWrapper):
-    for mm_idx, unet_idx in enumerate([1, 2, 4, 5, 7, 8, 10, 11]):
-        mm_idx0, mm_idx1 = mm_idx // 2, mm_idx % 2
-        unet.input_blocks[unet_idx].append(
-            motion_module.down_blocks[mm_idx0].motion_modules[mm_idx1]
-        )
-
-    for unet_idx in range(12):
-        mm_idx0, mm_idx1 = unet_idx // 3, unet_idx % 3
-        if unet_idx % 3 == 2 and unet_idx != 11:
-            unet.output_blocks[unet_idx].insert(
-                -1, motion_module.up_blocks[mm_idx0].motion_modules[mm_idx1]
-            )
-        else:
-            unet.output_blocks[unet_idx].append(
-                motion_module.up_blocks[mm_idx0].motion_modules[mm_idx1]
-            )
-    if motion_module.is_v2:
-        unet.middle_block.insert(-1, motion_module.mid_block.motion_modules[0])
-
-    unet.motion_module = motion_module
-
-
-def eject_motion_module_from_unet(unet):
-    for unet_idx in [1, 2, 4, 5, 7, 8, 10, 11]:
-        unet.input_blocks[unet_idx].pop(-1)
-
-    for unet_idx in range(12):
-        if unet_idx % 3 == 2 and unet_idx != 11:
-            unet.output_blocks[unet_idx].pop(-2)
-        else:
-            unet.output_blocks[unet_idx].pop(-1)
-
-    if unet.motion_module.is_v2:
-        unet.middle_block.pop(-2)
-
-    del unet.motion_module
-
-
-injectors = {
-    "legacy": inject_motion_module_to_unet_legacy,
-    "default": inject_motion_module_to_unet,
-}
-
-ejectors = {
-    "legacy": eject_motion_module_from_unet_legacy,
-    "default": eject_motion_module_from_unet,
-}
+video_formats_dir = os.path.join(os.path.dirname(os.path.abspath(__file__)), "..", "video_formats")
+video_formats = ["video/" + x[:-5] for x in os.listdir(video_formats_dir)]


 class AnimateDiffModuleLoader:
@@ -182,147 +30,97 @@ class AnimateDiffModuleLoader:
            "required": {
                "model_name": (get_available_models(),),
            },
+            "optional": {
+                "lora_stack": ("MOTION_LORA_STACK",),
+            },
        }

    RETURN_TYPES = ("MOTION_MODULE",)
    CATEGORY = "Animate Diff"
    FUNCTION = "load_motion_module"

+    def inject_loras(self, motion_module: MotionWrapper, lora_stack: List[Tuple[Dict[str, Tensor], float]]):
+        for lora in lora_stack:
+            (state_dict, alpha) = lora
+
+            for key in state_dict:
+                layer_infos = key.split(".")
+
+                curr_layer = motion_module
+                while len(layer_infos) > 0:
+                    temp_name = layer_infos.pop(0)
+                    curr_layer = curr_layer.__getattr__(temp_name)
+
+                curr_layer.weight.data += alpha * state_dict[key].to(curr_layer.weight.data.device)
+
+    def eject_loras(self, motion_module: MotionWrapper, lora_stack: List[Tuple[float, Dict[str, Tensor]]]):
+        lora_stack.reverse() # should not matter but just in case
+        for lora in lora_stack:
+            (state_dict, alpha) = lora
+
+            for key in state_dict:
+                layer_infos = key.split(".")
+
+                curr_layer = motion_module
+                while len(layer_infos) > 0:
+                    temp_name = layer_infos.pop(0)
+                    curr_layer = curr_layer.__getattr__(temp_name)
+
+                curr_layer.weight.data -= alpha * state_dict[key].to(curr_layer.weight.data.device)
+
    def load_motion_module(
        self,
        model_name: str,
+        lora_stack: List = None,
    ):
        motion_module = load_motion_module(model_name)

+        # inject loras
+        if motion_module.is_v2:
+            if hasattr(motion_module, "lora_stack") and isinstance(motion_module.lora_stack, list):
+                self.eject_loras(motion_module, motion_module.lora_stack)
+                delattr(motion_module, "lora_stack")
+
+            if isinstance(lora_stack, list):
+                self.inject_loras(motion_module, lora_stack)
+                setattr(motion_module, "lora_stack", lora_stack)
+
+        elif isinstance(lora_stack, list):
+            logger.warning("LoRA is provided but only motion module v2 is supported.")
+
        return (motion_module,)


-class AnimateDiffSampler(KSampler):
+class AnimateDiffLoraLoader:
    @classmethod
    def INPUT_TYPES(s):
-        inputs = {
+        return {
            "required": {
-                "motion_module": ("MOTION_MODULE",),
-                "inject_method": (["default", "legacy"],),
-                "frame_number": (
-                    "INT",
-                    {"default": 16, "min": 2, "max": 32, "step": 1},
-                ),
-            }
+                "lora_name": (get_available_loras(),),
+                "alpha": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 10.0, "step": 0.001}),
+            },
+            "optional": {
+                "lora_stack": ("MOTION_LORA_STACK",),
+            },
        }
-        inputs["required"].update(KSampler.INPUT_TYPES()["required"])
-        return inputs

-    FUNCTION = "animatediff_sample"
+    RETURN_TYPES = ("MOTION_LORA_STACK",)
    CATEGORY = "Animate Diff"
+    FUNCTION = "load_lora"

-    def __init__(self) -> None:
-        super().__init__()
-        self.prev_beta = None
-        self.prev_linear_start = None
-        self.prev_linear_end = None
-
-    def override_beta_schedule(self, model: BaseModel):
-        logger.info(f"Override beta schedule.")
-        self.prev_beta = model.get_buffer("betas").cpu().clone()
-        self.prev_linear_start = model.linear_start
-        self.prev_linear_end = model.linear_end
-        model.register_schedule(
-            given_betas=None,
-            beta_schedule="sqrt_linear",
-            timesteps=1000,
-            linear_start=0.00085,
-            linear_end=0.012,
-            cosine_s=8e-3,
-        )
-
-    def restore_beta_schedule(self, model: BaseModel):
-        logger.info(f"Restoring beta schedule.")
-        model.register_schedule(
-            given_betas=self.prev_beta,
-            linear_start=self.prev_linear_start,
-            linear_end=self.prev_linear_end,
-        )
-        self.prev_beta = None
-        self.prev_linear_start = None
-        self.prev_linear_end = None
-
-    def inject_motion_module(
-        self, model, motion_module: MotionWrapper, inject_method: str, frame_number: int
-    ):
-        model = model.clone()
-        unet = model.model.diffusion_model
-
-        logger.info(f"Injecting motion module with method {inject_method}.")
-        motion_module.set_video_length(frame_number)
-        injectors[inject_method](unet, motion_module)
-        self.override_beta_schedule(model.model)
-        if not motion_module.is_v2:
-            logger.info(f"Hacking GroupNorm.forward function.")
-            torch.nn.GroupNorm.forward = groupnorm_mm_factory(frame_number)
-
-        return model
-
-    def eject_motion_module(self, model, inject_method):
-        unet = model.model.diffusion_model
-
-        self.restore_beta_schedule(model.model)
-        if not unet.motion_module.is_v2:
-            logger.info(f"Restore GroupNorm.forward function.")
-            torch.nn.GroupNorm.forward = orig_groupnorm_forward
-
-        logger.info(f"Ejecting motion module with method {inject_method}.")
-        ejectors[inject_method](unet)
-
-    def animatediff_sample(
+    def load_lora(
        self,
-        motion_module,
-        inject_method,
-        frame_number,
-        model,
-        seed,
-        steps,
-        cfg,
-        sampler_name,
-        scheduler,
-        positive,
-        negative,
-        latent_image,
-        denoise=1.0,
+        lora_name: str,
+        alpha: float,
+        lora_stack: List = None,
    ):
-        model = self.inject_motion_module(
-            model, motion_module, inject_method, frame_number
-        )
+        if not lora_stack:
+            lora_stack = []

-        init_frames = len(latent_image["samples"])
-        samples = latent_image["samples"][:init_frames, :, :, :].clone().cpu()
+        lora = load_lora(lora_name)
+        lora_stack.append((lora, alpha))

-        if init_frames < frame_number:
-            last_frame = samples[-1].unsqueeze(0)
-            repeated_last_frames = last_frame.repeat(
-                frame_number - init_frames, 1, 1, 1
-            )
-            samples = torch.cat((samples, repeated_last_frames), dim=0)
-
-        latent_image = {"samples": samples}
-
-        try:
-            return super().sample(
-                model,
-                seed,
-                steps,
-                cfg,
-                sampler_name,
-                scheduler,
-                positive,
-                negative,
-                latent_image,
-                denoise=denoise,
-            )
-        except:
-            raise
-        finally:
-            self.eject_motion_module(model, inject_method)
+        return (lora_stack,)


 class AnimateDiffCombine:
@@ -336,11 +134,10 @@ class AnimateDiffCombine:
                    {"default": 8, "min": 1, "max": 24, "step": 1},
                ),
                "loop_count": ("INT", {"default": 0, "min": 0, "max": 100, "step": 1}),
-                "save_image": ([True, False],),
+                "save_image": ("BOOLEAN", {"default": True}),
                "filename_prefix": ("STRING", {"default": "animate_diff"}),
-                "format": (["image/gif", "image/webp"] +
-                           ["video/"+x[:-5] for x in folder_paths.get_filename_list("video_formats")],),
-                "pingpong": ([False, True],),
+                "format": (["image/gif", "image/webp"] + video_formats,),
+                "pingpong": ("BOOLEAN", {"default": False}),
            },
            "hidden": {
                "prompt": "PROMPT",
@@ -373,11 +170,7 @@ class AnimateDiffCombine:
            frames.append(img)

        # save image
-        output_dir = (
-            folder_paths.get_output_directory()
-            if save_image
-            else folder_paths.get_temp_directory()
-        )
+        output_dir = folder_paths.get_output_directory() if save_image else folder_paths.get_temp_directory()
        (
            full_output_folder,
            filename,
@@ -426,16 +219,31 @@ class AnimateDiffCombine:
            ffmpeg_path = shutil.which("ffmpeg")
            if ffmpeg_path is None:
                raise ProcessLookupError("Could not find ffmpeg")
-            video_format_path = folder_paths.get_full_path(
-                "video_formats", format_ext + ".json")
-            with open(video_format_path, 'r') as stream:
+            video_format_path = os.path.join(video_formats_dir, format_ext + ".json")
+            with open(video_format_path, "r") as stream:
                video_format = json.load(stream)
            file = f"{filename}_{counter:05}_.{video_format['extension']}"
            file_path = os.path.join(full_output_folder, file)
            dimensions = f"{frames[0].width}x{frames[0].height}"
-            args = [ffmpeg_path, "-v", "error", "-f", "rawvideo", "-pix_fmt", "rgb24",
-                    "-s", dimensions, "-r", str(frame_rate), "-i", "-"] \
-                + video_format['main_pass'] + [file_path]
+            args = (
+                [
+                    ffmpeg_path,
+                    "-v",
+                    "error",
+                    "-f",
+                    "rawvideo",
+                    "-pix_fmt",
+                    "rgb24",
+                    "-s",
+                    dimensions,
+                    "-r",
+                    str(frame_rate),
+                    "-i",
+                    "-",
+                ]
+                + video_format["main_pass"]
+                + [file_path]
+            )

            env = os.environ
            if "environment" in video_format:
@@ -462,17 +270,16 @@ class LoadVideo:
        if not os.path.exists(input_dir):
            os.makedirs(input_dir, exist_ok=True)

-        files = [f"video/{f}" for f in os.listdir(input_dir) if os.path.isfile(
-            os.path.join(input_dir, f))]
+        files = [f"video/{f}" for f in os.listdir(input_dir) if os.path.isfile(os.path.join(input_dir, f))]

        return {
            "required": {
                "video": (sorted(files), {"video_upload": True}),
            },
            "optional": {
-                "frame_start": ("INT", {"default": 0, "min": 0, "max": 0xffffffff, "step": 1}),
+                "frame_start": ("INT", {"default": 0, "min": 0, "max": 0xFFFFFFFF, "step": 1}),
                "frame_limit": ("INT", {"default": 16, "min": 1, "max": 10240, "step": 1}),
-            }
+            },
        }

    CATEGORY = "Animate Diff/Utils"
@@ -495,6 +302,7 @@ class LoadVideo:
        return frames

    def load_video(self, video_path, frame_start: int, frame_limit: int):
+        ensure_opencv()
        import cv2

        video = cv2.VideoCapture(video_path)
@@ -517,24 +325,23 @@ class LoadVideo:
        return frames

    def load(self, video: str, frame_start=0, frame_limit=16):
-        print("path", video)
        video_path = folder_paths.get_annotated_filepath(video)
        (_, ext) = os.path.splitext(video_path)

        if ext.lower() in {".gif", ".webp"}:
            frames = self.load_gif(video_path, frame_start, frame_limit)
-        elif ext.lower() in {".webp", ".mp4", ".mov", ".avi"}:
+        elif ext.lower() in {".webp", ".mp4", ".mov", ".avi", ".webm"}:
            frames = self.load_video(video_path, frame_start, frame_limit)
        else:
            raise ValueError(f"Unsupported video format: {ext}")

-        return (torch.cat(frames, dim=0),)
+        return (torch.cat(frames, dim=0), len(frames))

    @classmethod
    def IS_CHANGED(s, image, *args, **kwargs):
        image_path = folder_paths.get_annotated_filepath(image)
        m = hashlib.sha256()
-        with open(image_path, 'rb') as f:
+        with open(image_path, "rb") as f:
            m.update(f.read())
        return m.digest().hex()

@@ -572,7 +379,7 @@ class ImageChunking:
            "required": {
                "images": ("IMAGE",),
                "chunk_size": ("INT", {"default": 16, "min": 1, "max": 1024, "step": 1}),
-                "allow_remainder": ([True, False],),
+                "allow_remainder": ("BOOLEAN", {"default": True}),
            },
        }

@@ -584,29 +391,31 @@ class ImageChunking:
    def chunk(self, images: Tensor, chunk_size: int, allow_remainder: bool):
        # Check if tensor is divisible into chunks of chunk_size
        if images.shape[0] % chunk_size != 0 and not allow_remainder:
-            raise ValueError(
-                "Tensor's first dimension is not divisible by chunk size")
+            raise ValueError("Tensor's first dimension is not divisible by chunk size")

        # Use torch.chunk to divide the tensor
-        chunk_count = images.shape[0] // chunk_size + \
-            images.shape[0] % chunk_size
+        chunk_count = images.shape[0] // chunk_size + images.shape[0] % chunk_size

        print("chunk_count", chunk_count)
        chunks = torch.chunk(images, chunk_count, dim=0)

-        return (list(chunks), )
+        return (list(chunks),)


 NODE_CLASS_MAPPINGS = {
    "AnimateDiffModuleLoader": AnimateDiffModuleLoader,
+    "AnimateDiffLoraLoader": AnimateDiffLoraLoader,
    "AnimateDiffCombine": AnimateDiffCombine,
    "AnimateDiffSampler": AnimateDiffSampler,
+    "AnimateDiffSlidingWindowOptions": AnimateDiffSlidingWindowOptions,
    "LoadVideo": LoadVideo,
    "ImageSizeAndBatchSize": ImageSizeAndBatchSize,
 }
 NODE_DISPLAY_NAME_MAPPINGS = {
    "AnimateDiffModuleLoader": "Animate Diff Module Loader",
+    "AnimateDiffLoraLoader": "Animate Diff Lora Loader",
    "AnimateDiffSampler": "Animate Diff Sampler",
+    "AnimateDiffSlidingWindowOptions": "Sliding Window Options",
    "AnimateDiffCombine": "Animate Diff Combine",
    "LoadVideo": "Load Video",
    "ImageSizeAndBatchSize": "Get Image Size + Batch Size",
@@ -0,0 +1,316 @@
+import torch
+from torch import Tensor
+from torch.nn.functional import group_norm
+from einops import rearrange
+
+import comfy.ldm.modules.diffusionmodules.openaimodel as openaimodel
+import comfy.model_management as model_management
+from comfy.model_base import BaseModel
+from comfy.ldm.modules.attention import SpatialTransformer
+from nodes import KSampler
+
+from .logger import logger
+from .motion_module import MotionWrapper, VanillaTemporalModule
+from .sliding_schedule import ContextSchedules
+from .sliding_context_sampling import SlidingContext, inject_sampling_function, eject_sampling_function
+
+
+SLIDING_CONTEXT_LENGTH = 16
+
+
+def forward_timestep_embed(ts, x, emb, context=None, transformer_options={}, output_shape=None):
+    for layer in ts:
+        if isinstance(layer, openaimodel.TimestepBlock):
+            x = layer(x, emb)
+        elif isinstance(layer, VanillaTemporalModule):
+            x = layer(x, context)
+        elif isinstance(layer, SpatialTransformer):
+            x = layer(x, context, transformer_options)
+            transformer_options["current_index"] += 1
+        elif isinstance(layer, openaimodel.Upsample):
+            x = layer(x, output_shape=output_shape)
+        else:
+            x = layer(x)
+    return x
+
+
+def groupnorm_mm_factory(video_length: int):
+    def groupnorm_mm_forward(self, input: Tensor) -> Tensor:
+        # axes_factor normalizes batch based on total conds and unconds passed in batch;
+        # the conds and unconds per batch can change based on VRAM optimizations that may kick in
+        axes_factor = input.size(0) // video_length
+
+        input = rearrange(input, "(b f) c h w -> b c f h w", b=axes_factor)
+        input = group_norm(input, self.num_groups, self.weight, self.bias, self.eps)
+        input = rearrange(input, "b c f h w -> (b f) c h w", b=axes_factor)
+        return input
+
+    return groupnorm_mm_forward
+
+
+orig_forward_timestep_embed = openaimodel.forward_timestep_embed
+orig_maximum_batch_area = model_management.maximum_batch_area
+orig_groupnorm_forward = torch.nn.GroupNorm.forward
+
+
+def inject_motion_module_to_unet_legacy(unet, motion_module: MotionWrapper):
+    for mm_idx, unet_idx in enumerate([1, 2, 4, 5, 7, 8, 10, 11]):
+        mm_idx0, mm_idx1 = mm_idx // 2, mm_idx % 2
+        unet.input_blocks[unet_idx].append(motion_module.down_blocks[mm_idx0].motion_modules[mm_idx1])
+
+    for unet_idx in range(12):
+        mm_idx0, mm_idx1 = unet_idx // 3, unet_idx % 3
+        if unet_idx % 2 == 2:
+            unet.output_blocks[unet_idx].insert(-1, motion_module.up_blocks[mm_idx0].motion_modules[mm_idx1])
+        else:
+            unet.output_blocks[unet_idx].append(motion_module.up_blocks[mm_idx0].motion_modules[mm_idx1])
+    if motion_module.is_v2:
+        unet.middle_block.insert(-1, motion_module.mid_block.motion_modules[0])
+
+    unet.motion_module = motion_module
+
+
+def eject_motion_module_from_unet_legacy(unet):
+    for unet_idx in [1, 2, 4, 5, 7, 8, 10, 11]:
+        unet.input_blocks[unet_idx].pop(-1)
+
+    for unet_idx in range(12):
+        if unet_idx % 2 == 2:
+            unet.output_blocks[unet_idx].pop(-2)
+        else:
+            unet.output_blocks[unet_idx].pop(-1)
+
+    if unet.motion_module.is_v2:
+        unet.middle_block.pop(-2)
+
+    del unet.motion_module
+
+
+def inject_motion_module_to_unet(unet, motion_module: MotionWrapper):
+    for mm_idx, unet_idx in enumerate([1, 2, 4, 5, 7, 8, 10, 11]):
+        mm_idx0, mm_idx1 = mm_idx // 2, mm_idx % 2
+        unet.input_blocks[unet_idx].append(motion_module.down_blocks[mm_idx0].motion_modules[mm_idx1])
+
+    for unet_idx in range(12):
+        mm_idx0, mm_idx1 = unet_idx // 3, unet_idx % 3
+        if unet_idx % 3 == 2 and unet_idx != 11:
+            unet.output_blocks[unet_idx].insert(-1, motion_module.up_blocks[mm_idx0].motion_modules[mm_idx1])
+        else:
+            unet.output_blocks[unet_idx].append(motion_module.up_blocks[mm_idx0].motion_modules[mm_idx1])
+    if motion_module.is_v2:
+        unet.middle_block.insert(-1, motion_module.mid_block.motion_modules[0])
+
+    unet.motion_module = motion_module
+
+
+def eject_motion_module_from_unet(unet):
+    for unet_idx in [1, 2, 4, 5, 7, 8, 10, 11]:
+        unet.input_blocks[unet_idx].pop(-1)
+
+    for unet_idx in range(12):
+        if unet_idx % 3 == 2 and unet_idx != 11:
+            unet.output_blocks[unet_idx].pop(-2)
+        else:
+            unet.output_blocks[unet_idx].pop(-1)
+
+    if unet.motion_module.is_v2:
+        unet.middle_block.pop(-2)
+
+    del unet.motion_module
+
+
+injectors = {
+    "legacy": inject_motion_module_to_unet_legacy,
+    "default": inject_motion_module_to_unet,
+}
+
+ejectors = {
+    "legacy": eject_motion_module_from_unet_legacy,
+    "default": eject_motion_module_from_unet,
+}
+
+
+class AnimateDiffSlidingWindowOptions:
+    @classmethod
+    def INPUT_TYPES(s):
+        return {
+            "required": {
+                "context_length": ("INT", {"default": SLIDING_CONTEXT_LENGTH, "min": 2, "max": 32}),
+                "context_stride": ("INT", {"default": 1, "min": 1, "max": 32}),
+                "context_overlap": ("INT", {"default": 4, "min": 0, "max": 32}),
+                "context_schedule": (ContextSchedules.CONTEXT_SCHEDULE_LIST, {"default": ContextSchedules.UNIFORM}),
+                "closed_loop": ("BOOLEAN", {"default": False}),
+            }
+        }
+
+    RETURN_TYPES = ("SLIDING_WINDOW_OPTS",)
+    FUNCTION = "init_options"
+    CATEGORY = "Animate Diff"
+
+    def init_options(self, context_length, context_stride, context_overlap, context_schedule, closed_loop):
+        ctx = SlidingContext(
+            context_length=context_length,
+            context_stride=context_stride,
+            context_overlap=context_overlap,
+            context_schedule=context_schedule,
+            closed_loop=closed_loop,
+        )
+
+        return (ctx,)
+
+
+class AnimateDiffSampler(KSampler):
+    @classmethod
+    def INPUT_TYPES(s):
+        inputs = {
+            "required": {
+                "motion_module": ("MOTION_MODULE",),
+                "inject_method": (["default", "legacy"],),
+                "frame_number": (
+                    "INT",
+                    {"default": 16, "min": 2, "max": 10000, "step": 1},
+                ),
+            }
+        }
+        inputs["required"].update(KSampler.INPUT_TYPES()["required"])
+        inputs["optional"] = {"sliding_window_opts": ("SLIDING_WINDOW_OPTS",)}
+        return inputs
+
+    FUNCTION = "animatediff_sample"
+    CATEGORY = "Animate Diff"
+
+    def __init__(self) -> None:
+        super().__init__()
+        self.prev_beta = None
+        self.prev_linear_start = None
+        self.prev_linear_end = None
+
+    def override_beta_schedule(self, model: BaseModel):
+        self.prev_beta = model.get_buffer("betas").cpu().clone().detach()
+        self.prev_linear_start = model.linear_start
+        self.prev_linear_end = model.linear_end
+        model.register_schedule(
+            given_betas=None,
+            beta_schedule="sqrt_linear",
+            timesteps=1000,
+            linear_start=0.00085,
+            linear_end=0.012,
+            cosine_s=8e-3,
+        )
+
+    def restore_beta_schedule(self, model: BaseModel):
+        model.register_schedule(
+            given_betas=self.prev_beta,
+            linear_start=self.prev_linear_start,
+            linear_end=self.prev_linear_end,
+        )
+        self.prev_beta = None
+        self.prev_linear_start = None
+        self.prev_linear_end = None
+
+    def inject_motion_module(self, model, motion_module: MotionWrapper, inject_method: str, frame_number: int):
+        model = model.clone()
+        unet = model.model.diffusion_model
+
+        logger.info(f"Injecting motion module with method {inject_method}.")
+        motion_module.set_video_length(frame_number)
+        injectors[inject_method](unet, motion_module)
+        self.override_beta_schedule(model.model)
+        openaimodel.forward_timestep_embed = forward_timestep_embed
+        if not motion_module.is_v2:
+            logger.info(f"Hacking GroupNorm.forward function.")
+            torch.nn.GroupNorm.forward = groupnorm_mm_factory(frame_number)
+
+        return model
+
+    def inject_sliding_sampler(self, video_length, sliding_window_opts: SlidingContext = None):
+        ctx = sliding_window_opts.copy() if sliding_window_opts else SlidingContext()
+        ctx.video_length = video_length
+
+        inject_sampling_function(ctx)
+
+    def eject_motion_module(self, model, inject_method):
+        unet = model.model.diffusion_model
+
+        self.restore_beta_schedule(model.model)
+        openaimodel.forward_timestep_embed = orig_forward_timestep_embed
+        if not unet.motion_module.is_v2:
+            logger.info(f"Restore GroupNorm.forward function.")
+            torch.nn.GroupNorm.forward = orig_groupnorm_forward
+
+        logger.info(f"Ejecting motion module with method {inject_method}.")
+        ejectors[inject_method](unet)
+
+    def eject_sliding_sampler(self):
+        eject_sampling_function()
+
+    def animatediff_sample(
+        self,
+        motion_module,
+        inject_method,
+        frame_number,
+        model,
+        seed,
+        steps,
+        cfg,
+        sampler_name,
+        scheduler,
+        positive,
+        negative,
+        latent_image,
+        denoise=1.0,
+        sliding_window_opts: SlidingContext = None,
+        **kwargs,
+    ):
+        # init latents
+        samples = latent_image["samples"]
+        init_frames = len(samples)
+        if init_frames < frame_number:
+            # TODO: apply different noise to each frame
+            last_frame = samples[-1].clone().cpu().unsqueeze(0)
+            repeated_last_frames = last_frame.repeat(frame_number - init_frames, 1, 1, 1)
+            samples = torch.cat((samples, repeated_last_frames), dim=0)
+
+        latent_image = {"samples": samples}
+
+        # validate context_length
+        context_length = sliding_window_opts.context_length if sliding_window_opts else SLIDING_CONTEXT_LENGTH
+        is_sliding = frame_number > context_length
+        video_length = context_length if is_sliding else frame_number
+
+        if video_length > motion_module.encoding_max_len:
+            error = f'{"context_length" if is_sliding else "frame_number"} = {video_length}'
+            raise ValueError(
+                f"AnimateDiff model {motion_module.mm_type} has upper limit of {motion_module.encoding_max_len} frames, but received {error}."
+            )
+
+        # inject motion module
+        model = self.inject_motion_module(model, motion_module, inject_method, video_length)
+
+        # inject sliding sampler
+        if is_sliding:
+            self.inject_sliding_sampler(frame_number, sliding_window_opts=sliding_window_opts)
+
+        try:
+            return super().sample(
+                model,
+                seed,
+                steps,
+                cfg,
+                sampler_name,
+                scheduler,
+                positive,
+                negative,
+                latent_image,
+                denoise=denoise,
+                **kwargs,
+            )
+        except:
+            raise
+        finally:
+            # eject motion module
+            self.eject_motion_module(model, inject_method)
+
+            # eject sliding sampler
+            if is_sliding:
+                self.eject_sliding_sampler()
@@ -0,0 +1,436 @@
+import math
+import torch
+from torch import Tensor
+from typing import List, Dict
+
+import comfy.utils
+import comfy.sample
+import comfy.samplers as comfy_samplers
+import comfy.model_management as model_management
+from comfy.controlnet import ControlBase
+from comfy.model_patcher import ModelPatcher
+
+from .logger import logger
+from .sliding_schedule import get_context_scheduler, ContextSchedules
+
+
+orig_comfy_sample = comfy.sample.sample
+orig_sampling_function = comfy_samplers.sampling_function
+
+
+def lcm(a, b):
+    return abs(a * b) // math.gcd(a, b)
+
+
+class SlidingContext:
+    def __init__(
+        self,
+        context_length=16,
+        context_stride=1,
+        context_overlap=4,
+        context_schedule=ContextSchedules.UNIFORM,
+        closed_loop=False,
+        video_length=0,
+        current_step=0,
+        total_steps=0,
+    ):
+        self.context_length = context_length
+        self.context_stride = context_stride
+        self.context_overlap = context_overlap
+        self.context_schedule = context_schedule
+        self.closed_loop = closed_loop
+        self.video_length = video_length
+        self.current_step = current_step
+        self.total_steps = total_steps
+
+    def copy(self):
+        return SlidingContext(
+            context_length=self.context_length,
+            context_stride=self.context_stride,
+            context_overlap=self.context_overlap,
+            context_schedule=self.context_schedule,
+            closed_loop=self.closed_loop,
+            video_length=self.video_length,
+            current_step=self.current_step,
+            total_steps=self.total_steps,
+        )
+
+
+def __sliding_sample_factory(ctx: SlidingContext):
+    logger.info(f"Injecting sliding context sampling function.")
+    logger.info(f"Video length: {ctx.video_length}")
+    logger.info(f"Context length: {ctx.context_length}")
+    logger.info(f"Context schedule: {ctx.context_schedule}")
+
+    context_scheduler = get_context_scheduler(ctx.context_schedule)
+
+    def sample(model: ModelPatcher, *args, **kwargs):
+        orig_callback = kwargs.pop("callback", None)
+        start_step = kwargs.get("start_step") or 0
+
+        # adjust progressbar to account for context frames
+        def callback(step, x0, x, total_steps):
+            if orig_callback:
+                orig_callback(step, x0, x, total_steps)
+
+            ctx.current_step = start_step + step + 1
+
+        return orig_comfy_sample(model, *args, **kwargs, callback=callback)
+
+    def sampling_function(model_function, x, timestep, uncond, cond, cond_scale, model_options={}, seed=None):
+        def get_area_and_mult(conds, x_in, timestep_in):
+            area = (x_in.shape[2], x_in.shape[3], 0, 0)
+            strength = 1.0
+
+            if "timestep_start" in conds:
+                timestep_start = conds["timestep_start"]
+                if timestep_in[0] > timestep_start:
+                    return None
+            if "timestep_end" in conds:
+                timestep_end = conds["timestep_end"]
+                if timestep_in[0] < timestep_end:
+                    return None
+            if "area" in conds:
+                area = conds["area"]
+            if "strength" in conds:
+                strength = conds["strength"]
+
+            input_x = x_in[:, :, area[2] : area[0] + area[2], area[3] : area[1] + area[3]]
+            if "mask" in conds:
+                # Scale the mask to the size of the input
+                # The mask should have been resized as we began the sampling process
+                mask_strength = 1.0
+                if "mask_strength" in conds:
+                    mask_strength = conds["mask_strength"]
+                mask = conds["mask"]
+                assert mask.shape[1] == x_in.shape[2]
+                assert mask.shape[2] == x_in.shape[3]
+                mask = mask[:, area[2] : area[0] + area[2], area[3] : area[1] + area[3]] * mask_strength
+                mask = mask.unsqueeze(1).repeat(input_x.shape[0] // mask.shape[0], input_x.shape[1], 1, 1)
+            else:
+                mask = torch.ones_like(input_x)
+            mult = mask * strength
+
+            if "mask" not in conds:
+                rr = 8
+                if area[2] != 0:
+                    for t in range(rr):
+                        mult[:, :, t : 1 + t, :] *= (1.0 / rr) * (t + 1)
+                if (area[0] + area[2]) < x_in.shape[2]:
+                    for t in range(rr):
+                        mult[:, :, area[0] - 1 - t : area[0] - t, :] *= (1.0 / rr) * (t + 1)
+                if area[3] != 0:
+                    for t in range(rr):
+                        mult[:, :, :, t : 1 + t] *= (1.0 / rr) * (t + 1)
+                if (area[1] + area[3]) < x_in.shape[3]:
+                    for t in range(rr):
+                        mult[:, :, :, area[1] - 1 - t : area[1] - t] *= (1.0 / rr) * (t + 1)
+
+            conditionning = {}
+            model_conds = conds["model_conds"]
+            for c in model_conds:
+                conditionning[c] = model_conds[c].process_cond(batch_size=x_in.shape[0], device=x_in.device, area=area)
+
+            control = None
+            if "control" in conds:
+                control = conds["control"]
+
+            patches = None
+            if "gligen" in conds:
+                gligen = conds["gligen"]
+                patches = {}
+                gligen_type = gligen[0]
+                gligen_model = gligen[1]
+                if gligen_type == "position":
+                    gligen_patch = gligen_model.model.set_position(input_x.shape, gligen[2], input_x.device)
+                else:
+                    gligen_patch = gligen_model.model.set_empty(input_x.shape, input_x.device)
+
+                patches["middle_patch"] = [gligen_patch]
+
+            return (input_x, mult, conditionning, area, control, patches)
+
+        def cond_equal_size(c1, c2):
+            if c1 is c2:
+                return True
+            if c1.keys() != c2.keys():
+                return False
+            for k in c1:
+                if not c1[k].can_concat(c2[k]):
+                    return False
+            return True
+
+        def can_concat_cond(c1, c2):
+            if c1[0].shape != c2[0].shape:
+                return False
+
+            # control
+            if (c1[4] is None) != (c2[4] is None):
+                return False
+            if c1[4] is not None:
+                if c1[4] is not c2[4]:
+                    return False
+
+            # patches
+            if (c1[5] is None) != (c2[5] is None):
+                return False
+            if c1[5] is not None:
+                if c1[5] is not c2[5]:
+                    return False
+
+            return cond_equal_size(c1[2], c2[2])
+
+        def cond_cat(c_list):
+            c_crossattn = []
+            c_concat = []
+            c_adm = []
+            crossattn_max_len = 0
+
+            temp = {}
+            for x in c_list:
+                for k in x:
+                    cur = temp.get(k, [])
+                    cur.append(x[k])
+                    temp[k] = cur
+
+            out = {}
+            for k in temp:
+                conds = temp[k]
+                out[k] = conds[0].concat(conds[1:])
+
+            return out
+
+        def calc_cond_uncond_batch(model_function, cond, uncond, x_in, timestep, max_total_area, model_options):
+            out_cond = torch.zeros_like(x_in)
+            out_count = torch.ones_like(x_in) / 100000.0
+
+            out_uncond = torch.zeros_like(x_in)
+            out_uncond_count = torch.ones_like(x_in) / 100000.0
+
+            COND = 0
+            UNCOND = 1
+
+            to_run = []
+            for x in cond:
+                p = get_area_and_mult(x, x_in, timestep)
+                if p is None:
+                    continue
+
+                to_run += [(p, COND)]
+            if uncond is not None:
+                for x in uncond:
+                    p = get_area_and_mult(x, x_in, timestep)
+                    if p is None:
+                        continue
+
+                    to_run += [(p, UNCOND)]
+
+            while len(to_run) > 0:
+                first = to_run[0]
+                first_shape = first[0][0].shape
+                to_batch_temp = []
+                for x in range(len(to_run)):
+                    if can_concat_cond(to_run[x][0], first[0]):
+                        to_batch_temp += [x]
+
+                to_batch_temp.reverse()
+                to_batch = to_batch_temp[:1]
+
+                for i in range(1, len(to_batch_temp) + 1):
+                    batch_amount = to_batch_temp[: len(to_batch_temp) // i]
+                    if len(batch_amount) * first_shape[0] * first_shape[2] * first_shape[3] < max_total_area:
+                        to_batch = batch_amount
+                        break
+
+                input_x = []
+                mult = []
+                c = []
+                cond_or_uncond = []
+                area = []
+                control = None
+                patches = None
+                for x in to_batch:
+                    o = to_run.pop(x)
+                    p = o[0]
+                    input_x += [p[0]]
+                    mult += [p[1]]
+                    c += [p[2]]
+                    area += [p[3]]
+                    cond_or_uncond += [o[1]]
+                    control = p[4]
+                    patches = p[5]
+
+                batch_chunks = len(cond_or_uncond)
+                input_x = torch.cat(input_x)
+                c = cond_cat(c)
+                timestep_ = torch.cat([timestep] * batch_chunks)
+
+                if control is not None:
+                    c["control"] = control.get_control(input_x, timestep_, c, len(cond_or_uncond))
+
+                transformer_options = {}
+                if "transformer_options" in model_options:
+                    transformer_options = model_options["transformer_options"].copy()
+
+                if patches is not None:
+                    if "patches" in transformer_options:
+                        cur_patches = transformer_options["patches"].copy()
+                        for p in patches:
+                            if p in cur_patches:
+                                cur_patches[p] = cur_patches[p] + patches[p]
+                            else:
+                                cur_patches[p] = patches[p]
+                    else:
+                        transformer_options["patches"] = patches
+
+                transformer_options["cond_or_uncond"] = cond_or_uncond[:]
+                c["transformer_options"] = transformer_options
+
+                if "model_function_wrapper" in model_options:
+                    output = model_options["model_function_wrapper"](
+                        model_function,
+                        {"input": input_x, "timestep": timestep_, "c": c, "cond_or_uncond": cond_or_uncond},
+                    ).chunk(batch_chunks)
+                else:
+                    output = model_function(input_x, timestep_, **c).chunk(batch_chunks)
+                del input_x
+
+                for o in range(batch_chunks):
+                    if cond_or_uncond[o] == COND:
+                        out_cond[:, :, area[o][2] : area[o][0] + area[o][2], area[o][3] : area[o][1] + area[o][3]] += (
+                            output[o] * mult[o]
+                        )
+                        out_count[
+                            :, :, area[o][2] : area[o][0] + area[o][2], area[o][3] : area[o][1] + area[o][3]
+                        ] += mult[o]
+                    else:
+                        out_uncond[
+                            :, :, area[o][2] : area[o][0] + area[o][2], area[o][3] : area[o][1] + area[o][3]
+                        ] += (output[o] * mult[o])
+                        out_uncond_count[
+                            :, :, area[o][2] : area[o][0] + area[o][2], area[o][3] : area[o][1] + area[o][3]
+                        ] += mult[o]
+                del mult
+
+            out_cond /= out_count
+            del out_count
+            out_uncond /= out_uncond_count
+            del out_uncond_count
+
+            return out_cond, out_uncond
+
+        # sliding_calc_cond_uncond_batch inspired by ashen's initial hack for 16-frame sliding context:
+        # https://github.com/comfyanonymous/ComfyUI/compare/master...ashen-sensored:ComfyUI:master
+        def sliding_calc_cond_uncond_batch(
+            model_function, cond, uncond, x_in, timestep, max_total_area, model_options
+        ):
+            # figure out how input is split
+            axes_factor = x.size(0) // ctx.video_length
+
+            # prepare final cond, uncond, and out_count
+            cond_final = torch.zeros_like(x)
+            uncond_final = torch.zeros_like(x)
+            out_count_final = torch.zeros((x.shape[0], 1, 1, 1), device=x.device)
+
+            def prepare_control_objects(control: ControlBase, full_idxs: list[int]):
+                if control.previous_controlnet is not None:
+                    prepare_control_objects(control.previous_controlnet, full_idxs)
+                control.sub_idxs = full_idxs
+                control.full_latent_length = ctx.video_length
+                control.context_length = ctx.context_length
+
+            def get_resized_cond(cond_in: List[Dict], full_idxs) -> list:
+                # reuse or resize cond items to match context requirements
+                resized_cond = []
+                # cond object is a list containing a list - outer list is irrelevant, so just loop through it
+                for actual_cond in cond_in:
+                    new_cond_item = actual_cond.copy()
+                    for key, cond_item in new_cond_item.items():
+                        if isinstance(cond_item, Tensor):
+                            # check that tensor is the expected length - x.size(0)
+                            if cond_item.size(0) == x.size(0):
+                                # if so, it's subsetting time - tell controls the expected indeces so they can handle them
+                                actual_cond_item = cond_item[full_idxs]
+                                new_cond_item[key] = actual_cond_item
+                        elif key == "control":
+                            control_item = cond_item
+                            if hasattr(control_item, "sub_idxs"):
+                                prepare_control_objects(control_item, full_idxs)
+                            else:
+                                raise ValueError(
+                                    f"Control type {type(control_item).__name__} may not support required features for sliding context window; use Control objects from Kosinkadink/Advanced-ControlNet nodes."
+                                )
+                            new_cond_item[key] = cond_item
+                    resized_cond.append(new_cond_item)
+                return resized_cond
+
+            # perform calc_cond_uncond_batch per context window
+            for ctx_idxs in context_scheduler(
+                ctx.current_step,
+                ctx.total_steps,
+                ctx.video_length,
+                ctx.context_length,
+                ctx.context_stride,
+                ctx.context_overlap,
+                ctx.closed_loop,
+            ):
+                # account for all portions of input frames
+                full_idxs = []
+                for n in range(axes_factor):
+                    for ind in ctx_idxs:
+                        full_idxs.append((ctx.video_length * n) + ind)
+                # get subsections of x, timestep, cond, uncond, cond_concat
+                sub_x = x[full_idxs]
+                sub_timestep = timestep[full_idxs]
+                sub_cond = get_resized_cond(cond, full_idxs) if cond is not None else None
+                sub_uncond = get_resized_cond(uncond, full_idxs) if uncond is not None else None
+
+                sub_cond_out, sub_uncond_out = calc_cond_uncond_batch(
+                    model_function,
+                    sub_cond,
+                    sub_uncond,
+                    sub_x,
+                    sub_timestep,
+                    max_total_area,
+                    model_options,
+                )
+
+                cond_final[full_idxs] += sub_cond_out
+                uncond_final[full_idxs] += sub_uncond_out
+                out_count_final[full_idxs] += 1  # increment which indeces were used
+
+            # normalize cond and uncond via division by context usage counts
+            cond_final /= out_count_final
+            uncond_final /= out_count_final
+            return cond_final, uncond_final
+
+        max_total_area = model_management.maximum_batch_area()
+        if math.isclose(cond_scale, 1.0):
+            uncond = None
+
+        cond, uncond = sliding_calc_cond_uncond_batch(
+            model_function, cond, uncond, x, timestep, max_total_area, model_options
+        )
+
+        if "sampler_cfg_function" in model_options:
+            args = {"cond": cond, "uncond": uncond, "cond_scale": cond_scale, "timestep": timestep}
+            return model_options["sampler_cfg_function"](args)
+        else:
+            return uncond + (cond - uncond) * cond_scale
+
+    return (sample, sampling_function)
+
+
+def inject_sampling_function(ctx: SlidingContext):
+    global orig_comfy_sample, orig_sampling_function
+    orig_comfy_sample = comfy.sample.sample
+    orig_sampling_function = comfy_samplers.sampling_function
+
+    (sample, sampling_function) = __sliding_sample_factory(ctx)
+    comfy.sample.sample = sample
+    comfy_samplers.sampling_function = sampling_function
+
+
+def eject_sampling_function():
+    comfy.sample.sample = orig_comfy_sample
+    comfy_samplers.sampling_function = orig_sampling_function
@@ -0,0 +1,155 @@
+# from https://github.com/neggles/animatediff-cli/blob/main/src/animatediff/pipelines/context.py
+from typing import Callable, Optional
+
+import numpy as np
+
+
+class ContextSchedules:
+    UNIFORM = "uniform"
+    UNIFORM_CONSTANT = "uniform_constant"
+    UNIFORM_V2 = "uniform v2"
+
+    CONTEXT_SCHEDULE_LIST = [UNIFORM]
+
+
+# Returns fraction that has denominator that is a power of 2
+def ordered_halving(val, print_final=False):
+    # get binary value, padded with 0s for 64 bits
+    bin_str = f"{val:064b}"
+    # flip binary value, padding included
+    bin_flip = bin_str[::-1]
+    # convert binary to int
+    as_int = int(bin_flip, 2)
+    # divide by 1 << 64, equivalent to 2**64, or 18446744073709551616,
+    # or b10000000000000000000000000000000000000000000000000000000000000000 (1 with 64 zero's)
+    final = as_int / (1 << 64)
+    if print_final:
+        print(f"$$$$ final: {final}")
+    return final
+
+
+# Generator that returns lists of latent indeces to diffuse on
+def uniform(
+    step: int = ...,
+    num_steps: Optional[int] = None,
+    num_frames: int = ...,
+    context_size: Optional[int] = None,
+    context_stride: int = 3,
+    context_overlap: int = 4,
+    closed_loop: bool = True,
+    print_final: bool = False,
+):
+    if num_frames <= context_size:
+        yield list(range(num_frames))
+        return
+
+    context_stride = min(context_stride, int(np.ceil(np.log2(num_frames / context_size))) + 1)
+
+    for context_step in 1 << np.arange(context_stride):
+        pad = int(round(num_frames * ordered_halving(step, print_final)))
+        for j in range(
+            int(ordered_halving(step) * context_step) + pad,
+            num_frames + pad + (0 if closed_loop else -context_overlap),
+            (context_size * context_step - context_overlap),
+        ):
+            yield [e % num_frames for e in range(j, j + context_size * context_step, context_step)]
+
+
+def uniform_v2(
+    step: int = ...,
+    num_steps: Optional[int] = None,
+    num_frames: int = ...,
+    context_size: Optional[int] = None,
+    context_stride: int = 3,
+    context_overlap: int = 4,
+    closed_loop: bool = True,
+    print_final: bool = False,
+):
+    if num_frames <= context_size:
+        yield list(range(num_frames))
+        return
+
+    context_stride = min(context_stride, int(np.ceil(np.log2(num_frames / context_size))) + 1)
+
+    pad = int(round(num_frames * ordered_halving(step, print_final)))
+    for context_step in 1 << np.arange(context_stride):
+        j_initial = int(ordered_halving(step) * context_step) + pad
+        for j in range(
+            j_initial,
+            num_frames + pad - context_overlap,
+            (context_size * context_step - context_overlap),
+        ):
+            if context_size * context_step > num_frames:
+                # On the final context_step,
+                # ensure no frame appears in the window twice
+                yield [e % num_frames for e in range(j, j + num_frames, context_step)]
+                continue
+            j = j % num_frames
+            if j > (j + context_size * context_step) % num_frames and not closed_loop:
+                yield [e for e in range(j, num_frames, context_step)]
+                j_stop = (j + context_size * context_step) % num_frames
+                # When  ((num_frames % (context_size - context_overlap)+context_overlap) % context_size != 0,
+                # This can cause 'superflous' runs where all frames in
+                # a context window have already been processed during
+                # the first context window of this stride and step.
+                # While the following commented if should prevent this,
+                # I believe leaving it in is more correct as it maintains
+                # the total conditional passes per frame over a large total steps
+                # if j_stop > context_overlap:
+                yield [e for e in range(0, j_stop, context_step)]
+                continue
+            yield [e % num_frames for e in range(j, j + context_size * context_step, context_step)]
+
+
+def uniform_constant(
+    step: int = ...,
+    num_steps: Optional[int] = None,
+    num_frames: int = ...,
+    context_size: Optional[int] = None,
+    context_stride: int = 3,
+    context_overlap: int = 4,
+    closed_loop: bool = True,
+    print_final: bool = False,
+):
+    if num_frames <= context_size:
+        yield list(range(num_frames))
+        return
+
+    context_stride = min(context_stride, int(np.ceil(np.log2(num_frames / context_size))) + 1)
+
+    # want to avoid loops that connect end to beginning
+
+    for context_step in 1 << np.arange(context_stride):
+        pad = int(round(num_frames * ordered_halving(step, print_final)))
+        for j in range(
+            int(ordered_halving(step) * context_step) + pad,
+            num_frames + pad + (0 if closed_loop else -context_overlap),
+            (context_size * context_step - context_overlap),
+        ):
+            skip_this_window = False
+            prev_val = -1
+            to_yield = []
+            for e in range(j, j + context_size * context_step, context_step):
+                e = e % num_frames
+                # if not a closed loop and loops back on itself, should be skipped
+                if not closed_loop and e < prev_val:
+                    skip_this_window = True
+                    break
+                to_yield.append(e)
+                prev_val = e
+            if skip_this_window:
+                continue
+            # yield if not skipped
+            yield to_yield
+
+
+def get_context_scheduler(name: str) -> Callable:
+    match name:
+        case ContextSchedules.UNIFORM:
+            return uniform
+        case ContextSchedules.UNIFORM_CONSTANT:
+            return uniform_constant
+        case ContextSchedules.UNIFORM_V2:
+            return uniform_v2
+        case _:
+            raise ValueError(f"Unknown context_overlap policy {name}")
@@ -1,13 +1,32 @@
+import sys
 import torch
 import numpy as np
+import subprocess
 from PIL import Image

+
+from .logger import logger
+
 # Tensor to PIL
 def tensor2pil(image):
-    return Image.fromarray(
-        np.clip(255.0 * image.cpu().numpy().squeeze(), 0, 255).astype(np.uint8)
-    )
+    return Image.fromarray(np.clip(255.0 * image.cpu().numpy().squeeze(), 0, 255).astype(np.uint8))
+

 # Convert PIL to Tensor
 def pil2tensor(image):
    return torch.from_numpy(np.array(image).astype(np.float32) / 255.0).unsqueeze(0)
+
+
+def ensure_opencv():
+    if "python_embeded" in sys.executable or "python_embedded" in sys.executable:
+        pip_install = [sys.executable, "-s", "-m", "pip", "install"]
+    else:
+        pip_install = [sys.executable, "-m", "pip", "install"]
+
+    try:
+        import cv2
+    except Exception as e:
+        try:
+            subprocess.check_call(pip_install + ['opencv-python'])
+        except:
+            logger.error(f"Failed to install 'opencv-python'. Please, install manually.")
@@ -0,0 +1 @@
+opencv-python
@@ -738,10 +738,10 @@
        60,
        140
      ],
-      "size": [
-        340,
-        110
-      ],
+      "size": {
+        "0": 340,
+        "1": 110
+      },
      "flags": {},
      "order": 10,
      "mode": 0,
@@ -780,7 +780,7 @@
      ],
      "size": {
        "0": 310,
-        "1": 330
+        "1": 350
      },
      "flags": {},
      "order": 30,
@@ -812,6 +812,11 @@
          "name": "latent_image",
          "type": "LATENT",
          "link": 80
+        },
+        {
+          "name": "sliding_window_opts",
+          "type": "SLIDING_WINDOW_OPTS",
+          "link": null
        }
      ],
      "outputs": [
@@ -1039,10 +1044,10 @@
        2140,
        140
      ],
-      "size": [
-        360,
-        552
-      ],
+      "size": {
+        "0": 360,
+        "1": 552
+      },
      "flags": {},
      "order": 32,
      "mode": 0,
@@ -1070,8 +1075,7 @@
        true,
        "AnimateDiff",
        "image/gif",
-        true,
-        "/view?filename=AnimateDiff_00090_.gif&subfolder=&type=output&format=image%2Fgif"
+        true
      ]
    },
    {
@@ -1110,49 +1114,6 @@
      "color": "#1a572e",
      "bgcolor": "#2e6b42"
    },
-    {
-      "id": 44,
-      "type": "VAEDecode",
-      "pos": [
-        1900,
-        520
-      ],
-      "size": {
-        "0": 210,
-        "1": 46
-      },
-      "flags": {},
-      "order": 31,
-      "mode": 0,
-      "inputs": [
-        {
-          "name": "samples",
-          "type": "LATENT",
-          "link": 81
-        },
-        {
-          "name": "vae",
-          "type": "VAE",
-          "link": 82
-        }
-      ],
-      "outputs": [
-        {
-          "name": "IMAGE",
-          "type": "IMAGE",
-          "links": [
-            172
-          ],
-          "shape": 3,
-          "slot_index": 0
-        }
-      ],
-      "properties": {
-        "Node name for S&R": "VAEDecode"
-      },
-      "color": "#2e571a",
-      "bgcolor": "#426b2e"
-    },
    {
      "id": 4,
      "type": "CheckpointLoaderSimple",
@@ -1208,10 +1169,10 @@
        420,
        530
      ],
-      "size": [
-        260,
-        170
-      ],
+      "size": {
+        "0": 260,
+        "1": 170
+      },
      "flags": {},
      "order": 28,
      "mode": 0,
@@ -1478,6 +1439,49 @@
      ],
      "color": "#1a572e",
      "bgcolor": "#2e6b42"
+    },
+    {
+      "id": 44,
+      "type": "VAEDecode",
+      "pos": [
+        1898,
+        541
+      ],
+      "size": {
+        "0": 210,
+        "1": 46
+      },
+      "flags": {},
+      "order": 31,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "samples",
+          "type": "LATENT",
+          "link": 81
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 82
+        }
+      ],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [
+            172
+          ],
+          "shape": 3,
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "VAEDecode"
+      },
+      "color": "#2e571a",
+      "bgcolor": "#426b2e"
    }
  ],
  "links": [
@@ -1,6 +1,6 @@
 {
-  "last_node_id": 106,
-  "last_link_id": 189,
+  "last_node_id": 107,
+  "last_link_id": 199,
  "nodes": [
    {
      "id": 16,
@@ -21,7 +21,7 @@
          "name": "MOTION_MODULE",
          "type": "MOTION_MODULE",
          "links": [
-            78
+            193
          ],
          "shape": 3,
          "slot_index": 0
@@ -77,10 +77,10 @@
        1240,
        140
      ],
-      "size": [
-        360,
-        732
-      ],
+      "size": {
+        "0": 360,
+        "1": 732
+      },
      "flags": {},
      "order": 13,
      "mode": 0,
@@ -108,8 +108,7 @@
        true,
        "AnimateDiff",
        "image/gif",
-        true,
-        "/view?filename=AnimateDiff_00092_.gif&subfolder=&type=output&format=image%2Fgif"
+        true
      ]
    },
    {
@@ -131,7 +130,7 @@
          "name": "MODEL",
          "type": "MODEL",
          "links": [
-            79
+            194
          ],
          "slot_index": 0
        },
@@ -167,10 +166,10 @@
        60,
        300
      ],
-      "size": [
-        310,
-        100
-      ],
+      "size": {
+        "0": 310,
+        "1": 100
+      },
      "flags": {},
      "order": 6,
      "mode": 0,
@@ -207,10 +206,10 @@
        60,
        140
      ],
-      "size": [
-        310,
-        110
-      ],
+      "size": {
+        "0": 310,
+        "1": 110
+      },
      "flags": {},
      "order": 5,
      "mode": 0,
@@ -240,94 +239,6 @@
      "color": "#572e1a",
      "bgcolor": "#6b422e"
    },
-    {
-      "id": 41,
-      "type": "AnimateDiffSampler",
-      "pos": [
-        900,
-        140
-      ],
-      "size": [
-        310,
-        330
-      ],
-      "flags": {},
-      "order": 11,
-      "mode": 0,
-      "inputs": [
-        {
-          "name": "motion_module",
-          "type": "MOTION_MODULE",
-          "link": 78,
-          "slot_index": 0
-        },
-        {
-          "name": "model",
-          "type": "MODEL",
-          "link": 79,
-          "slot_index": 1
-        },
-        {
-          "name": "positive",
-          "type": "CONDITIONING",
-          "link": 176
-        },
-        {
-          "name": "negative",
-          "type": "CONDITIONING",
-          "link": 180
-        },
-        {
-          "name": "latent_image",
-          "type": "LATENT",
-          "link": 80
-        },
-        {
-          "name": "frame_number",
-          "type": "INT",
-          "link": 185,
-          "widget": {
-            "name": "frame_number",
-            "config": [
-              "INT",
-              {
-                "default": 16,
-                "min": 2,
-                "max": 32,
-                "step": 1
-              }
-            ]
-          }
-        }
-      ],
-      "outputs": [
-        {
-          "name": "LATENT",
-          "type": "LATENT",
-          "links": [
-            81
-          ],
-          "shape": 3,
-          "slot_index": 0
-        }
-      ],
-      "properties": {
-        "Node name for S&R": "AnimateDiffSampler"
-      },
-      "widgets_values": [
-        "default",
-        16,
-        345029849956754,
-        "fixed",
-        20,
-        8,
-        "euler",
-        "normal",
-        1
-      ],
-      "color": "#57571a",
-      "bgcolor": "#6b6b2e"
-    },
    {
      "id": 39,
      "type": "ControlNetApplyAdvanced",
@@ -335,10 +246,10 @@
        471,
        275
      ],
-      "size": [
-        300,
-        170
-      ],
+      "size": {
+        "0": 300,
+        "1": 170
+      },
      "flags": {},
      "order": 9,
      "mode": 0,
@@ -369,7 +280,7 @@
          "name": "positive",
          "type": "CONDITIONING",
          "links": [
-            176
+            195
          ],
          "shape": 3,
          "slot_index": 0
@@ -378,7 +289,7 @@
          "name": "negative",
          "type": "CONDITIONING",
          "links": [
-            180
+            196
          ],
          "shape": 3,
          "slot_index": 1
@@ -395,50 +306,6 @@
      "color": "#43571a",
      "bgcolor": "#576b2e"
    },
-    {
-      "id": 44,
-      "type": "VAEDecode",
-      "pos": [
-        1000,
-        520
-      ],
-      "size": {
-        "0": 210,
-        "1": 46
-      },
-      "flags": {},
-      "order": 12,
-      "mode": 0,
-      "inputs": [
-        {
-          "name": "samples",
-          "type": "LATENT",
-          "link": 81
-        },
-        {
-          "name": "vae",
-          "type": "VAE",
-          "link": 82
-        }
-      ],
-      "outputs": [
-        {
-          "name": "IMAGE",
-          "type": "IMAGE",
-          "links": [
-            172,
-            187
-          ],
-          "shape": 3,
-          "slot_index": 0
-        }
-      ],
-      "properties": {
-        "Node name for S&R": "VAEDecode"
-      },
-      "color": "#2e571a",
-      "bgcolor": "#426b2e"
-    },
    {
      "id": 103,
      "type": "LoadVideo",
@@ -490,10 +357,10 @@
        520,
        630
      ],
-      "size": [
-        210,
-        80
-      ],
+      "size": {
+        "0": 210,
+        "1": 80
+      },
      "flags": {},
      "order": 10,
      "mode": 0,
@@ -501,7 +368,7 @@
        {
          "name": "width",
          "type": "INT",
-          "link": 189,
+          "link": 190,
          "widget": {
            "name": "width",
            "config": [
@@ -518,7 +385,7 @@
        {
          "name": "height",
          "type": "INT",
-          "link": 188,
+          "link": 191,
          "widget": {
            "name": "height",
            "config": [
@@ -538,7 +405,7 @@
          "name": "LATENT",
          "type": "LATENT",
          "links": [
-            80
+            197
          ],
          "shape": 3,
          "slot_index": 0
@@ -555,62 +422,6 @@
      "color": "#1a572e",
      "bgcolor": "#2e6b42"
    },
-    {
-      "id": 104,
-      "type": "ImageSizeAndBatchSize",
-      "pos": [
-        300,
-        630
-      ],
-      "size": [
-        190,
-        80
-      ],
-      "flags": {},
-      "order": 7,
-      "mode": 0,
-      "inputs": [
-        {
-          "name": "image",
-          "type": "IMAGE",
-          "link": 182
-        }
-      ],
-      "outputs": [
-        {
-          "name": "width",
-          "type": "INT",
-          "links": [
-            188
-          ],
-          "shape": 3,
-          "slot_index": 0
-        },
-        {
-          "name": "height",
-          "type": "INT",
-          "links": [
-            189
-          ],
-          "shape": 3,
-          "slot_index": 1
-        },
-        {
-          "name": "batch_size",
-          "type": "INT",
-          "links": [
-            185
-          ],
-          "shape": 3,
-          "slot_index": 2
-        }
-      ],
-      "properties": {
-        "Node name for S&R": "ImageSizeAndBatchSize"
-      },
-      "color": "#1a5757",
-      "bgcolor": "#2e6b6b"
-    },
    {
      "id": 36,
      "type": "ControlNetLoaderAdvanced",
@@ -618,10 +429,10 @@
        -280,
        540
      ],
-      "size": [
-        310,
-        60
-      ],
+      "size": {
+        "0": 310,
+        "1": 60
+      },
      "flags": {},
      "order": 4,
      "mode": 0,
@@ -660,10 +471,10 @@
        70,
        830
      ],
-      "size": [
-        530,
-        420
-      ],
+      "size": {
+        "0": 530,
+        "1": 420
+      },
      "flags": {},
      "order": 8,
      "mode": 0,
@@ -687,10 +498,10 @@
        670,
        830
      ],
-      "size": [
-        530,
-        420
-      ],
+      "size": {
+        "0": 530,
+        "1": 420
+      },
      "flags": {},
      "order": 14,
      "mode": 0,
@@ -706,6 +517,195 @@
      },
      "color": "#1a5757",
      "bgcolor": "#2e6b6b"
+    },
+    {
+      "id": 104,
+      "type": "ImageSizeAndBatchSize",
+      "pos": [
+        258,
+        631
+      ],
+      "size": {
+        "0": 226.8000030517578,
+        "1": 80
+      },
+      "flags": {},
+      "order": 7,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "image",
+          "type": "IMAGE",
+          "link": 182
+        }
+      ],
+      "outputs": [
+        {
+          "name": "width",
+          "type": "INT",
+          "links": [
+            190
+          ],
+          "shape": 3,
+          "slot_index": 0
+        },
+        {
+          "name": "height",
+          "type": "INT",
+          "links": [
+            191
+          ],
+          "shape": 3,
+          "slot_index": 1
+        },
+        {
+          "name": "batch_size",
+          "type": "INT",
+          "links": [
+            198
+          ],
+          "shape": 3,
+          "slot_index": 2
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "ImageSizeAndBatchSize"
+      },
+      "color": "#1a5757",
+      "bgcolor": "#2e6b6b"
+    },
+    {
+      "id": 44,
+      "type": "VAEDecode",
+      "pos": [
+        1000,
+        548
+      ],
+      "size": {
+        "0": 210,
+        "1": 46
+      },
+      "flags": {},
+      "order": 12,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "samples",
+          "type": "LATENT",
+          "link": 199
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 82
+        }
+      ],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [
+            172,
+            187
+          ],
+          "shape": 3,
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "VAEDecode"
+      },
+      "color": "#2e571a",
+      "bgcolor": "#426b2e"
+    },
+    {
+      "id": 107,
+      "type": "AnimateDiffSampler",
+      "pos": [
+        881,
+        141
+      ],
+      "size": [
+        330,
+        350
+      ],
+      "flags": {},
+      "order": 11,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "motion_module",
+          "type": "MOTION_MODULE",
+          "link": 193
+        },
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 194
+        },
+        {
+          "name": "positive",
+          "type": "CONDITIONING",
+          "link": 195
+        },
+        {
+          "name": "negative",
+          "type": "CONDITIONING",
+          "link": 196
+        },
+        {
+          "name": "latent_image",
+          "type": "LATENT",
+          "link": 197
+        },
+        {
+          "name": "sliding_window_opts",
+          "type": "SLIDING_WINDOW_OPTS",
+          "link": null
+        },
+        {
+          "name": "frame_number",
+          "type": "INT",
+          "link": 198,
+          "widget": {
+            "name": "frame_number",
+            "config": [
+              "INT",
+              {
+                "default": 16,
+                "min": 2,
+                "max": 10000,
+                "step": 1
+              }
+            ]
+          }
+        }
+      ],
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": [
+            199
+          ],
+          "shape": 3,
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "AnimateDiffSampler"
+      },
+      "widgets_values": [
+        "default",
+        16,
+        0,
+        "randomize",
+        20,
+        8,
+        "euler",
+        "normal",
+        1
+      ]
    }
  ],
  "links": [
@@ -749,38 +749,6 @@
      1,
      "CONDITIONING"
    ],
-    [
-      78,
-      16,
-      0,
-      41,
-      0,
-      "MOTION_MODULE"
-    ],
-    [
-      79,
-      4,
-      0,
-      41,
-      1,
-      "MODEL"
-    ],
-    [
-      80,
-      20,
-      0,
-      41,
-      4,
-      "LATENT"
-    ],
-    [
-      81,
-      41,
-      0,
-      44,
-      0,
-      "LATENT"
-    ],
    [
      82,
      13,
@@ -797,22 +765,6 @@
      0,
      "IMAGE"
    ],
-    [
-      176,
-      39,
-      0,
-      41,
-      2,
-      "CONDITIONING"
-    ],
-    [
-      180,
-      39,
-      1,
-      41,
-      3,
-      "CONDITIONING"
-    ],
    [
      181,
      103,
@@ -829,14 +781,6 @@
      0,
      "IMAGE"
    ],
-    [
-      185,
-      104,
-      2,
-      41,
-      5,
-      "INT"
-    ],
    [
      186,
      103,
@@ -854,20 +798,76 @@
      "IMAGE"
    ],
    [
-      188,
+      190,
      104,
      0,
      20,
+      0,
+      "INT"
+    ],
+    [
+      191,
+      104,
+      1,
+      20,
      1,
      "INT"
    ],
    [
-      189,
-      104,
+      193,
+      16,
+      0,
+      107,
+      0,
+      "MOTION_MODULE"
+    ],
+    [
+      194,
+      4,
+      0,
+      107,
      1,
+      "MODEL"
+    ],
+    [
+      195,
+      39,
+      0,
+      107,
+      2,
+      "CONDITIONING"
+    ],
+    [
+      196,
+      39,
+      1,
+      107,
+      3,
+      "CONDITIONING"
+    ],
+    [
+      197,
      20,
      0,
+      107,
+      4,
+      "LATENT"
+    ],
+    [
+      198,
+      104,
+      2,
+      107,
+      6,
      "INT"
+    ],
+    [
+      199,
+      107,
+      0,
+      44,
+      0,
+      "LATENT"
    ]
  ],
  "groups": [],
@@ -195,10 +195,10 @@
        571,
        712
      ],
-      "size": [
-        275.35137939453125,
-        82
-      ],
+      "size": {
+        "0": 275.35137939453125,
+        "1": 82
+      },
      "flags": {},
      "order": 11,
      "mode": 0,
@@ -315,7 +315,7 @@
      ],
      "size": {
        "0": 315,
-        "1": 330
+        "1": 350
      },
      "flags": {},
      "order": 12,
@@ -347,6 +347,11 @@
          "name": "latent_image",
          "type": "LATENT",
          "link": 47
+        },
+        {
+          "name": "sliding_window_opts",
+          "type": "SLIDING_WINDOW_OPTS",
+          "link": null
        }
      ],
      "outputs": [
@@ -424,7 +429,7 @@
      ],
      "size": {
        "0": 315,
-        "1": 330
+        "1": 350
      },
      "flags": {},
      "order": 6,
@@ -456,6 +461,11 @@
          "name": "latent_image",
          "type": "LATENT",
          "link": 35
+        },
+        {
+          "name": "sliding_window_opts",
+          "type": "SLIDING_WINDOW_OPTS",
+          "link": null
        }
      ],
      "outputs": [
@@ -518,48 +528,6 @@
        "klF8Anime2.ckpt"
      ]
    },
-    {
-      "id": 12,
-      "type": "AnimateDiffCombine",
-      "pos": [
-        1504,
-        481
-      ],
-      "size": [
-        325.7265625,
-        517.7265625
-      ],
-      "flags": {},
-      "order": 14,
-      "mode": 0,
-      "inputs": [
-        {
-          "name": "images",
-          "type": "IMAGE",
-          "link": 19
-        }
-      ],
-      "outputs": [
-        {
-          "name": "GIF",
-          "type": "GIF",
-          "links": null,
-          "shape": 3
-        }
-      ],
-      "properties": {
-        "Node name for S&R": "AnimateDiffCombine"
-      },
-      "widgets_values": [
-        8,
-        0,
-        true,
-        "AnimateDiff",
-        "image/webp",
-        false,
-        "/view?filename=AnimateDiff_00033_.webp&subfolder=&type=output&format=image%2Fwebp"
-      ]
-    },
    {
      "id": 27,
      "type": "VAEDecode",
@@ -604,13 +572,13 @@
      "id": 28,
      "type": "AnimateDiffCombine",
      "pos": [
-        1506,
-        -60
-      ],
-      "size": [
-        321.19170831298857,
-        513.0408935546875
+        1505,
+        -72
      ],
+      "size": {
+        "0": 321.19171142578125,
+        "1": 513.0408935546875
+      },
      "flags": {},
      "order": 10,
      "mode": 0,
@@ -637,9 +605,49 @@
        0,
        true,
        "AnimateDiff",
-        "image/webp",
-        false,
-        "/view?filename=AnimateDiff_00032_.webp&subfolder=&type=output&format=image%2Fwebp"
+        "image/gif",
+        false
+      ]
+    },
+    {
+      "id": 12,
+      "type": "AnimateDiffCombine",
+      "pos": [
+        1504,
+        481
+      ],
+      "size": {
+        "0": 325.7265625,
+        "1": 517.7265625
+      },
+      "flags": {},
+      "order": 14,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "link": 19
+        }
+      ],
+      "outputs": [
+        {
+          "name": "GIF",
+          "type": "GIF",
+          "links": null,
+          "shape": 3
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "AnimateDiffCombine"
+      },
+      "widgets_values": [
+        8,
+        0,
+        true,
+        "AnimateDiff",
+        "image/gif",
+        false
      ]
    }
  ],
@@ -0,0 +1,515 @@
+{
+  "last_node_id": 21,
+  "last_link_id": 38,
+  "nodes": [
+    {
+      "id": 6,
+      "type": "CLIPTextEncode",
+      "pos": [
+        415,
+        186
+      ],
+      "size": {
+        "0": 422.84503173828125,
+        "1": 164.31304931640625
+      },
+      "flags": {},
+      "order": 4,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 3
+        }
+      ],
+      "outputs": [
+        {
+          "name": "CONDITIONING",
+          "type": "CONDITIONING",
+          "links": [
+            29
+          ],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "CLIPTextEncode"
+      },
+      "widgets_values": [
+        "photo of coastline, rocks, storm weather, wind, waves, lightning, 8k uhd, dslr, soft lighting, high quality, film grain, Fujifilm XT3"
+      ]
+    },
+    {
+      "id": 8,
+      "type": "VAEDecode",
+      "pos": [
+        1253,
+        191
+      ],
+      "size": {
+        "0": 210,
+        "1": 46
+      },
+      "flags": {},
+      "order": 8,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "samples",
+          "type": "LATENT",
+          "link": 28
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 20
+        }
+      ],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [
+            19
+          ],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "VAEDecode"
+      }
+    },
+    {
+      "id": 12,
+      "type": "AnimateDiffCombine",
+      "pos": [
+        1254,
+        290
+      ],
+      "size": [
+        315,
+        507
+      ],
+      "flags": {},
+      "order": 9,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "link": 19
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "AnimateDiffCombine"
+      },
+      "widgets_values": [
+        8,
+        0,
+        false,
+        "AnimateDiff",
+        "image/gif",
+        false,
+        "/view?filename=AnimateDiff_00003_.gif&subfolder=&type=temp&format=image%2Fgif"
+      ]
+    },
+    {
+      "id": 7,
+      "type": "CLIPTextEncode",
+      "pos": [
+        413,
+        389
+      ],
+      "size": {
+        "0": 425.27801513671875,
+        "1": 180.6060791015625
+      },
+      "flags": {},
+      "order": 5,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 5
+        }
+      ],
+      "outputs": [
+        {
+          "name": "CONDITIONING",
+          "type": "CONDITIONING",
+          "links": [
+            30
+          ],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "CLIPTextEncode"
+      },
+      "widgets_values": [
+        "blur, haze, deformed iris, deformed pupils, semi-realistic, cgi, 3d, render, sketch, cartoon, drawing, anime, mutated hands and fingers, deformed, distorted, disfigured, poorly drawn, bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs, disconnected limbs, mutation, mutated, ugly, disgusting, amputation"
+      ]
+    },
+    {
+      "id": 20,
+      "type": "EmptyLatentImage",
+      "pos": [
+        522,
+        621
+      ],
+      "size": {
+        "0": 315,
+        "1": 106
+      },
+      "flags": {},
+      "order": 0,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": [
+            35
+          ],
+          "shape": 3,
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "EmptyLatentImage"
+      },
+      "widgets_values": [
+        512,
+        512,
+        1
+      ]
+    },
+    {
+      "id": 13,
+      "type": "VAELoader",
+      "pos": [
+        28,
+        223
+      ],
+      "size": {
+        "0": 315,
+        "1": 58
+      },
+      "flags": {},
+      "order": 1,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "VAE",
+          "type": "VAE",
+          "links": [
+            20
+          ],
+          "shape": 3,
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "VAELoader"
+      },
+      "widgets_values": [
+        "vae-ft-mse-840000-ema-pruned.safetensors"
+      ]
+    },
+    {
+      "id": 15,
+      "type": "AnimateDiffSampler",
+      "pos": [
+        882,
+        192
+      ],
+      "size": {
+        "0": 315,
+        "1": 350
+      },
+      "flags": {},
+      "order": 7,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "motion_module",
+          "type": "MOTION_MODULE",
+          "link": 24,
+          "slot_index": 0
+        },
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 25,
+          "slot_index": 1
+        },
+        {
+          "name": "positive",
+          "type": "CONDITIONING",
+          "link": 29
+        },
+        {
+          "name": "negative",
+          "type": "CONDITIONING",
+          "link": 30
+        },
+        {
+          "name": "latent_image",
+          "type": "LATENT",
+          "link": 35
+        },
+        {
+          "name": "sliding_window_opts",
+          "type": "SLIDING_WINDOW_OPTS",
+          "link": null
+        }
+      ],
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": [
+            28
+          ],
+          "shape": 3,
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "AnimateDiffSampler"
+      },
+      "widgets_values": [
+        "default",
+        14,
+        45987230,
+        "fixed",
+        25,
+        7.5,
+        "ddim",
+        "ddim_uniform",
+        1
+      ]
+    },
+    {
+      "id": 16,
+      "type": "AnimateDiffModuleLoader",
+      "pos": [
+        27,
+        345
+      ],
+      "size": {
+        "0": 315,
+        "1": 58
+      },
+      "flags": {},
+      "order": 6,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "lora_stack",
+          "type": "MOTION_LORA_STACK",
+          "link": 38,
+          "slot_index": 0
+        }
+      ],
+      "outputs": [
+        {
+          "name": "MOTION_MODULE",
+          "type": "MOTION_MODULE",
+          "links": [
+            24
+          ],
+          "shape": 3
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "AnimateDiffModuleLoader"
+      },
+      "widgets_values": [
+        "mm_sd_v15_v2.ckpt"
+      ]
+    },
+    {
+      "id": 21,
+      "type": "AnimateDiffLoraLoader",
+      "pos": [
+        -317,
+        350
+      ],
+      "size": [
+        310,
+        80
+      ],
+      "flags": {},
+      "order": 3,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "lora_stack",
+          "type": "MOTION_LORA_STACK",
+          "link": null
+        }
+      ],
+      "outputs": [
+        {
+          "name": "MOTION_LORA_STACK",
+          "type": "MOTION_LORA_STACK",
+          "links": [
+            38
+          ],
+          "shape": 3,
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "AnimateDiffLoraLoader"
+      },
+      "widgets_values": [
+        "v2_lora_ZoomIn.ckpt",
+        1
+      ]
+    },
+    {
+      "id": 4,
+      "type": "CheckpointLoaderSimple",
+      "pos": [
+        28,
+        457
+      ],
+      "size": {
+        "0": 315,
+        "1": 98
+      },
+      "flags": {},
+      "order": 2,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "MODEL",
+          "type": "MODEL",
+          "links": [
+            25
+          ],
+          "slot_index": 0
+        },
+        {
+          "name": "CLIP",
+          "type": "CLIP",
+          "links": [
+            3,
+            5
+          ],
+          "slot_index": 1
+        },
+        {
+          "name": "VAE",
+          "type": "VAE",
+          "links": [],
+          "slot_index": 2
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "CheckpointLoaderSimple"
+      },
+      "widgets_values": [
+        "RealisticVision_v20.safetensors"
+      ]
+    }
+  ],
+  "links": [
+    [
+      3,
+      4,
+      1,
+      6,
+      0,
+      "CLIP"
+    ],
+    [
+      5,
+      4,
+      1,
+      7,
+      0,
+      "CLIP"
+    ],
+    [
+      19,
+      8,
+      0,
+      12,
+      0,
+      "IMAGE"
+    ],
+    [
+      20,
+      13,
+      0,
+      8,
+      1,
+      "VAE"
+    ],
+    [
+      24,
+      16,
+      0,
+      15,
+      0,
+      "MOTION_MODULE"
+    ],
+    [
+      25,
+      4,
+      0,
+      15,
+      1,
+      "MODEL"
+    ],
+    [
+      28,
+      15,
+      0,
+      8,
+      0,
+      "LATENT"
+    ],
+    [
+      29,
+      6,
+      0,
+      15,
+      2,
+      "CONDITIONING"
+    ],
+    [
+      30,
+      7,
+      0,
+      15,
+      3,
+      "CONDITIONING"
+    ],
+    [
+      35,
+      20,
+      0,
+      15,
+      4,
+      "LATENT"
+    ],
+    [
+      38,
+      21,
+      0,
+      16,
+      0,
+      "MOTION_LORA_STACK"
+    ]
+  ],
+  "groups": [],
+  "config": {},
+  "extra": {},
+  "version": 0.4
+}
@@ -0,0 +1,502 @@
+{
+  "last_node_id": 21,
+  "last_link_id": 36,
+  "nodes": [
+    {
+      "id": 6,
+      "type": "CLIPTextEncode",
+      "pos": [
+        415,
+        186
+      ],
+      "size": {
+        "0": 422.84503173828125,
+        "1": 164.31304931640625
+      },
+      "flags": {},
+      "order": 5,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 3
+        }
+      ],
+      "outputs": [
+        {
+          "name": "CONDITIONING",
+          "type": "CONDITIONING",
+          "links": [
+            29
+          ],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "CLIPTextEncode"
+      },
+      "widgets_values": [
+        "masterpiece, best quality, 1girl, solo, cherry blossoms, hanami, pink flower, white flower, spring season, wisteria, petals, flower, plum blossoms, outdoors, falling petals, white hair, black eyes"
+      ]
+    },
+    {
+      "id": 8,
+      "type": "VAEDecode",
+      "pos": [
+        1253,
+        191
+      ],
+      "size": {
+        "0": 210,
+        "1": 46
+      },
+      "flags": {},
+      "order": 8,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "samples",
+          "type": "LATENT",
+          "link": 28
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 20
+        }
+      ],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [
+            19
+          ],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "VAEDecode"
+      }
+    },
+    {
+      "id": 12,
+      "type": "AnimateDiffCombine",
+      "pos": [
+        1254,
+        290
+      ],
+      "size": {
+        "0": 315,
+        "1": 342
+      },
+      "flags": {},
+      "order": 9,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "link": 19
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "AnimateDiffCombine"
+      },
+      "widgets_values": [
+        8,
+        0,
+        false,
+        "AnimateDiff",
+        "image/gif",
+        false
+      ]
+    },
+    {
+      "id": 16,
+      "type": "AnimateDiffModuleLoader",
+      "pos": [
+        27,
+        345
+      ],
+      "size": {
+        "0": 315,
+        "1": 58
+      },
+      "flags": {},
+      "order": 0,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "MOTION_MODULE",
+          "type": "MOTION_MODULE",
+          "links": [
+            24
+          ],
+          "shape": 3
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "AnimateDiffModuleLoader"
+      },
+      "widgets_values": [
+        "mm-Stabilized_mid.pth"
+      ]
+    },
+    {
+      "id": 4,
+      "type": "CheckpointLoaderSimple",
+      "pos": [
+        26,
+        474
+      ],
+      "size": {
+        "0": 315,
+        "1": 98
+      },
+      "flags": {},
+      "order": 1,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "MODEL",
+          "type": "MODEL",
+          "links": [
+            25
+          ],
+          "slot_index": 0
+        },
+        {
+          "name": "CLIP",
+          "type": "CLIP",
+          "links": [
+            3,
+            5
+          ],
+          "slot_index": 1
+        },
+        {
+          "name": "VAE",
+          "type": "VAE",
+          "links": [],
+          "slot_index": 2
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "CheckpointLoaderSimple"
+      },
+      "widgets_values": [
+        "AnimeLike25D_v11.safetensors"
+      ]
+    },
+    {
+      "id": 13,
+      "type": "VAELoader",
+      "pos": [
+        28,
+        223
+      ],
+      "size": {
+        "0": 315,
+        "1": 58
+      },
+      "flags": {},
+      "order": 2,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "VAE",
+          "type": "VAE",
+          "links": [
+            20
+          ],
+          "shape": 3,
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "VAELoader"
+      },
+      "widgets_values": [
+        "klF8Anime2.ckpt"
+      ]
+    },
+    {
+      "id": 7,
+      "type": "CLIPTextEncode",
+      "pos": [
+        413,
+        389
+      ],
+      "size": {
+        "0": 425.27801513671875,
+        "1": 180.6060791015625
+      },
+      "flags": {},
+      "order": 6,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 5
+        }
+      ],
+      "outputs": [
+        {
+          "name": "CONDITIONING",
+          "type": "CONDITIONING",
+          "links": [
+            30
+          ],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "CLIPTextEncode"
+      },
+      "widgets_values": [
+        "embedding:easynegative, embedding:badhandv4, "
+      ]
+    },
+    {
+      "id": 20,
+      "type": "EmptyLatentImage",
+      "pos": [
+        522,
+        621
+      ],
+      "size": {
+        "0": 315,
+        "1": 106
+      },
+      "flags": {},
+      "order": 3,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": [
+            35
+          ],
+          "shape": 3,
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "EmptyLatentImage"
+      },
+      "widgets_values": [
+        512,
+        512,
+        1
+      ]
+    },
+    {
+      "id": 21,
+      "type": "AnimateDiffSlidingWindowOptions",
+      "pos": [
+        517,
+        -34
+      ],
+      "size": {
+        "0": 315,
+        "1": 154
+      },
+      "flags": {},
+      "order": 4,
+      "mode": 0,
+      "outputs": [
+        {
+          "name": "SLIDING_WINDOW_OPTS",
+          "type": "SLIDING_WINDOW_OPTS",
+          "links": [
+            36
+          ],
+          "shape": 3
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "AnimateDiffSlidingWindowOptions"
+      },
+      "widgets_values": [
+        16,
+        1,
+        4,
+        "uniform",
+        true
+      ]
+    },
+    {
+      "id": 15,
+      "type": "AnimateDiffSampler",
+      "pos": [
+        882,
+        192
+      ],
+      "size": {
+        "0": 315,
+        "1": 350
+      },
+      "flags": {},
+      "order": 7,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "motion_module",
+          "type": "MOTION_MODULE",
+          "link": 24,
+          "slot_index": 0
+        },
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 25,
+          "slot_index": 1
+        },
+        {
+          "name": "positive",
+          "type": "CONDITIONING",
+          "link": 29
+        },
+        {
+          "name": "negative",
+          "type": "CONDITIONING",
+          "link": 30
+        },
+        {
+          "name": "latent_image",
+          "type": "LATENT",
+          "link": 35
+        },
+        {
+          "name": "sliding_window_opts",
+          "type": "SLIDING_WINDOW_OPTS",
+          "link": 36,
+          "slot_index": 5
+        }
+      ],
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": [
+            28
+          ],
+          "shape": 3,
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "AnimateDiffSampler"
+      },
+      "widgets_values": [
+        "default",
+        40,
+        345029849956677,
+        "fixed",
+        20,
+        8,
+        "euler",
+        "normal",
+        0.8
+      ]
+    }
+  ],
+  "links": [
+    [
+      3,
+      4,
+      1,
+      6,
+      0,
+      "CLIP"
+    ],
+    [
+      5,
+      4,
+      1,
+      7,
+      0,
+      "CLIP"
+    ],
+    [
+      19,
+      8,
+      0,
+      12,
+      0,
+      "IMAGE"
+    ],
+    [
+      20,
+      13,
+      0,
+      8,
+      1,
+      "VAE"
+    ],
+    [
+      24,
+      16,
+      0,
+      15,
+      0,
+      "MOTION_MODULE"
+    ],
+    [
+      25,
+      4,
+      0,
+      15,
+      1,
+      "MODEL"
+    ],
+    [
+      28,
+      15,
+      0,
+      8,
+      0,
+      "LATENT"
+    ],
+    [
+      29,
+      6,
+      0,
+      15,
+      2,
+      "CONDITIONING"
+    ],
+    [
+      30,
+      7,
+      0,
+      15,
+      3,
+      "CONDITIONING"
+    ],
+    [
+      35,
+      20,
+      0,
+      15,
+      4,
+      "LATENT"
+    ],
+    [
+      36,
+      21,
+      0,
+      15,
+      5,
+      "SLIDING_WINDOW_OPTS"
+    ]
+  ],
+  "groups": [],
+  "config": {},
+  "extra": {},
+  "version": 0.4
+}
Author	SHA1	Message	Date
Tung Nguyen	370a8e29f3	update get_resized_cond to match new cond format	2023-10-25 22:11:13 +07:00
Tung Nguyen	49916eae86	update sampling function to match new ComfyUI update	2023-10-25 16:30:06 +07:00
Tung Nguyen	dccfb2326e	fix: new comfyui break sampling	2023-10-25 15:24:50 +07:00
Tung Nguyen	62569f0184	fix VanillaTemporalModule.forward() missing 1 required positional argument: 'encoder_hidden_states'	2023-10-12 21:46:57 +07:00
Tung Nguyen	dc6268aea5	update to match new changes from ComfyUI	2023-10-12 21:44:21 +07:00
Tung Nguyen	9d32153349	increase max alpha for motion lora	2023-09-25 22:59:21 +07:00
Tung Nguyen (Blockchain)	63c068c59f	Support motion LoRA (#38 ) Support new motion LoRa from AnimateDiff	2023-09-25 22:53:44 +07:00
Tung Nguyen	a0bdb7e06c	add requirements.txt and auto install opencv-python when needed	2023-09-25 10:59:28 +07:00
Tung Nguyen	5d1d909b53	remove the buggy uniform_v2 context schedule	2023-09-23 04:57:17 +07:00
Tung Nguyen	52609c02a9	conform to use_split_cross_attention arg	2023-09-22 10:02:13 +07:00
Tung Nguyen	1aa6948063	using sub quadratic optimization when xformers is enabled	2023-09-22 04:14:01 +07:00
Tung Nguyen	71ac33b8f3	revert default context schedule to uniform	2023-09-21 15:06:41 +07:00
Tung Nguyen	8bd67a69cf	update README with more info on sliding window	2023-09-21 13:23:22 +07:00
Tung Nguyen	9f9513c236	use uniform_v2 as default context schedule	2023-09-21 13:12:47 +07:00
Tung Nguyen	149a6bb3de	update README with xformers issue	2023-09-21 05:50:10 +07:00
ArtVenture	3b8c80ba6e	Sliding window (#32 ) * sliding window feature * move all injections code to sampling time * remove video_formats from comfy folder_paths * update README & add example workflow	2023-09-21 05:40:29 +07:00
Tung Nguyen	c41af1c324	update example workflow to match new node updates	2023-09-21 05:39:19 +07:00