Merge pull request #27 from AustinMroz/upstream_video_format

ffmpeg improvements: webm quality, and additional video formats

README.md

@@ -5,20 +5,47 @@
 ## How to Use
 
 1. Clone this repo into `custom_nodes` folder.
-2. Download motion modules from [Google Drive](https://drive.google.com/drive/folders/1EqLC65eR1-W-sGD0Im7fkED6c8GkiNFI) | [HuggingFace](https://huggingface.co/guoyww/animatediff) | [CivitAI](https://civitai.com/models/108836) | [Baidu NetDisk](https://pan.baidu.com/s/18ZpcSM6poBqxWNHtnyMcxg?pwd=et8y). You only need to download one of `mm_sd_v14.ckpt` | `mm_sd_v15.ckpt`. Put the model weights under `comfyui-animatediff/models/`. DO NOT change model filename.
+2. Download motion modules and put them under `comfyui-animatediff/models/`.
+  * Original modules: [Google Drive](https://drive.google.com/drive/folders/1EqLC65eR1-W-sGD0Im7fkED6c8GkiNFI) | [HuggingFace](https://huggingface.co/guoyww/animatediff) | [CivitAI](https://civitai.com/models/108836) | [Baidu NetDisk](https://pan.baidu.com/s/18ZpcSM6poBqxWNHtnyMcxg?pwd=et8y)
+  * 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)
+
+## Nodes
+
+#### AnimateDiffLoader
+
+<img width="370" alt="image" src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/9d756d01-ea45-4d1c-8e48-56f2725c7ca1">
+
+#### AnimateDiffSampler
+- Mostly the same with `KSampler`
+- Use `AnimateDiffLoader` to load the motion module
+- `inject_method`: should left default
+- `frame_number`: animation length
+- `latent_image`: You can pass an `EmptyLatentImage`
+
+<img width="370" alt="image" src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/f22d6b36-ce36-44cc-80e8-dffe6f77b296">
+
+#### AnimateDiffCombine
+- Combine GIF frames and produce the GIF image
+- `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)
+
+<img width="370" alt="image" src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/381c5acc-06ef-43da-ada0-3dc76f37a3e4">
+
+#### Example Workflow
+
+<img width="1311" alt="image" src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/b7164539-bc58-4ef9-b178-d914e833805e">
+
+
+Workflow file: https://github.com/ArtVentureX/comfyui-animatediff/blob/main/workflow.json
 
 ## Samples
 
-### txt2img
+![23b44c29-29e8-4f48-ab3c-4df87c90c13f](https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/97efb96f-3d3d-4976-8789-78b88f89b2eb)
 
-<img width="1254" alt="ComfyUI AnimateDiff Usage" src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/a88e2141-c55f-4bdb-b6ca-9155b6639114">
+![25f6c60c-f8ac-4abe-984f-1559c355d7f6](https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/c39b26f7-a2af-4dc4-902f-c363e2e6f39a)
-
-![AnimateDiff_00001](https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/e48f148a-886b-4a0d-b589-9fa795b06936)
-
-### img2img
-<img width="1121" alt="Screenshot 2023-07-22 at 22 08 00" src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/600f96b0-df21-4437-917f-7eda35ab6363">
-
-![AnimateDiff_00002](https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/c78d64b9-b308-41ec-9804-bbde654d0b47)
 
 ## Known Issues
 
@@ -26,30 +53,14 @@
 
 ![AnimateDiff_00007_](https://github.com/ArtVentureX/comfyui-animatediff/assets/8894763/e6cd53cb-9878-45da-a58a-a15851882386)
 
-This is usually due to memory (VRAM) is not enough to process the whole image batch at the same time. Try reduce the image size and frame number.
+Work around:
 
-### GIF has Wartermark after update to the latest version
+- Shorter your prompt and negative prompt
+- Reduce resolution. AnimateDiff is trained on 512x512 images so it works best with 512x512 output.
+- Disable xformers with `--disable-xformers`
 
-See https://github.com/continue-revolution/sd-webui-animatediff/issues/31
+### GIF has Wartermark (especially when using mm_sd_v15)
 
-As mentioned in the issue thread, it seems to be due to the training dataset. The new version is the correct implementation and produces smoother GIFs compared to the older version.
+See: https://github.com/continue-revolution/sd-webui-animatediff/issues/31
 
-<table  class="center">
+Training data used by the authors of the AnimateDiff paper contained Shutterstock watermarks. Since mm_sd_v15 was finetuned on finer, less drastic movement, the motion module attempts to replicate the transparency of that watermark and does not get blurred away like mm_sd_v14. Try other community finetuned modules.
-    <tr>
-    <td>Old revision</td>
-    <td>New revision</td>
-    </tr>
-    <tr>
-    <td><img src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/8f1a6233-875f-4f0c-aa60-ba93e73b7d64" /></td>
-    <td><img src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/a2029eba-f519-437c-a0b5-1f881e099a20" /></td>
-    </tr>
-    <tr>
-    <td><img src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/41ec449f-1955-466c-bd38-6f2a55d654f8" /></td>
-    <td><img src="https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/766c2891-5d27-4052-99f9-be9862620919" /></td>
-    </tr>
-</table>
-
-
-I played around with both version and found that the watermark only present in some models, not always. So I've brought back the old method and also created a new node with the new method. You can try both to find the best fit for each model.
-
-![Screenshot 2023-07-28 at 18 14 14](https://github.com/ArtVentureX/comfyui-animatediff/assets/133728487/25cf6092-3e67-435e-86cc-43614ca7d6aa)

__init__.py

@@ -5,4 +5,6 @@ from .animatediff.model_utils import get_available_models
 if len(get_available_models()) == 0:
     logger.error("No models available. Please download one and put it in models folder")
 
-__all__ = ["NODE_CLASS_MAPPINGS", "NODE_DISPLAY_NAME_MAPPINGS"]
+WEB_DIRECTORY = "./web"
+
+__all__ = ["NODE_CLASS_MAPPINGS", "NODE_DISPLAY_NAME_MAPPINGS", "WEB_DIRECTORY"]

animatediff/model_utils.py

@@ -3,8 +3,6 @@ import hashlib
 
 import folder_paths
 
-from .logger import logger
-
 
 folder_paths.folder_names_and_paths["AnimateDiff"] = (
     [
@@ -13,17 +11,12 @@ folder_paths.folder_names_and_paths["AnimateDiff"] = (
     ],
     folder_paths.supported_pt_extensions,
 )
-
+folder_paths.folder_names_and_paths["video_formats"] = (
-known_models = {
+    [
-    "aa7fd8a200a89031edd84487e2a757c5315460eca528fa70d4b3885c399bffd5": "mm_sd_v14.ckpt",
+        os.path.join(os.path.dirname(os.path.abspath(__file__)), "..", "video_formats"),
-    "cf16ea656cb16124990c8e2c70a29c793f9841f3a2223073fac8bd89ebd9b69a": "mm_sd_v15.ckpt",
+    ],
-    "0aaf157b9c51a0ae07cb5d9ea7c51299f07bddc6f52025e1f9bb81cd763631df": "mm-Stabilized_high.pth",
+    [".json"]
-    "39de8b71b1c09f10f4602f5d585d82771a60d3cf282ba90215993e06afdfe875": "mm-Stabilized_mid.pth",
+)
-    "3cb569f7ce3dc6a10aa8438e666265cb9be3120d8f205de6a456acf46b6c99f4": "temporaldiff-v1-animatediff.ckpt",
-    "69ed0f5fef82b110aca51bcab73b21104242bc65d6ab4b8b2a2a94d31cad1bf0": "mm_sd_v15_v2.ckpt",
-}
-
-v2_models = ["69ed0f5fef82b110aca51bcab73b21104242bc65d6ab4b8b2a2a94d31cad1bf0"]
 
 
 def get_available_models():
@@ -34,25 +27,7 @@ def get_model_path(model_name):
     return folder_paths.get_full_path("AnimateDiff", model_name)
 
 
-def sha256_file(file_path):
+def get_model_hash(file_path):
     with open(file_path, "rb") as f:
         bytes = f.read()  # read entire file as bytes
         return hashlib.sha256(bytes).hexdigest()
-
-
-def validate_mm_model(model_name):
-    model_path = get_model_path(model_name)
-    model_hash = sha256_file(model_path)
-
-    if model_hash in known_models:
-        logger.info(f"You are using {model_name}, which has been tested and supported.")
-    else:
-        logger.warn(
-            f"Your model {model_name} has not been tested and supported."
-            "Either your download is incomplete or your model has not been tested. "
-            "Please use at your own risk."
-        )
-
-    using_v2 = model_hash in v2_models
-
-    return (model_hash, using_v2)

animatediff/motion_module.py

@@ -1,12 +1,11 @@
+import os
 import torch
-import torch.nn.functional as F
+from torch import Tensor, nn
-from torch import nn
 
 import math
 from einops import rearrange, repeat
 
-from comfy.ldm.modules.attention import FeedForward
+from comfy.ldm.modules.attention import FeedForward, CrossAttention
-from .attention_processor import Attention as CrossAttention
 
 
 def zero_module(module):
@@ -15,41 +14,106 @@ def zero_module(module):
         p.detach().zero_()
     return module
 
+
+# Merge from https://github.com/Kosinkadink/ComfyUI-AnimateDiff-Evolved
+def get_encoding_max_len(mm_state_dict: dict[str, Tensor]) -> int:
+    # use pos_encoder.pe entries to determine max length - [1, {max_length}, {320|640|1280}]
+    for key in mm_state_dict.keys():
+        if key.endswith("pos_encoder.pe"):
+            return mm_state_dict[key].size(1)  # get middle dim
+    raise ValueError(f"No pos_encoder.pe found in mm_state_dict")
+
+
+def has_mid_block(mm_state_dict: dict[str, Tensor]):
+    # check if keys contain mid_block
+    for key in mm_state_dict.keys():
+        if key.startswith("mid_block."):
+            return True
+    return False
+
+
 class MotionWrapper(nn.Module):
-    def __init__(self, mm_hash, is_v2 = False):
+    def __init__(self, mm_type: str, encoding_max_len: int = 24, is_v2=False):
         super().__init__()
-        if is_v2:
+        self.mm_type = mm_type
-            max_len = 32
+        self.is_v2 = is_v2
-        else:
-            max_len = 24
 
         self.down_blocks = nn.ModuleList([])
         self.up_blocks = nn.ModuleList([])
+        self.mid_block = None
+
         for c in (320, 640, 1280, 1280):
-            self.down_blocks.append(MotionModule(c, max_len=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, is_up=True, max_len=max_len))
+            self.up_blocks.append(
+                MotionModule(c, BlockType.UP, encoding_max_len=encoding_max_len)
+            )
         if is_v2:
-            self.mid_block = MotionModule(1280, max_len=max_len, is_mid=is_v2)
+            self.mid_block = MotionModule(
-        self.mm_hash = mm_hash
+                1280, BlockType.MID, encoding_max_len=encoding_max_len
-        self.is_v2 = is_v2
+            )
+
+    @classmethod
+    def from_pretrained(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)
+
+        mm = cls(mm_type, encoding_max_len=encoding_max_len, is_v2=is_v2)
+        mm.load_state_dict(mm_state_dict, strict=False)
+        return mm
+
+    def set_video_length(self, video_length: int):
+        for block in self.down_blocks:
+            block.set_video_length(video_length)
+        for block in self.up_blocks:
+            block.set_video_length(video_length)
+        if self.mid_block is not None:
+            self.mid_block.set_video_length(video_length)
+
+
+class BlockType:
+    UP = "up"
+    DOWN = "down"
+    MID = "mid"
 
 
 class MotionModule(nn.Module):
-    def __init__(self, in_channels, is_up=False, is_mid=False, max_len=24):
+    def __init__(
+        self,
+        in_channels,
+        block_type: BlockType,
+        encoding_max_len=24,
+    ):
         super().__init__()
-        if is_mid:
+        self.block_type = block_type
-            self.motion_modules = nn.ModuleList([get_motion_module(in_channels, max_len)])
+
+        if block_type == BlockType.MID:
+            self.motion_modules = nn.ModuleList(
+                [get_motion_module(in_channels, encoding_max_len)]
+            )
         else:
             self.motion_modules = nn.ModuleList(
-                [get_motion_module(in_channels, max_len), get_motion_module(in_channels, max_len)]
+                [
+                    get_motion_module(in_channels, encoding_max_len),
+                    get_motion_module(in_channels, encoding_max_len),
+                ]
             )
-            if is_up:
+            if block_type == BlockType.UP:
-                self.motion_modules.append(get_motion_module(in_channels, 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:
+            motion_module.set_video_length(video_length)
 
 
 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):
@@ -85,8 +149,13 @@ class VanillaTemporalModule(nn.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)
+        return self.temporal_transformer(
+            input_tensor, encoder_hidden_states, attention_mask
+        )
 
 
 class TemporalTransformer3DModel(nn.Module):
@@ -140,10 +209,12 @@ class TemporalTransformer3DModel(nn.Module):
             ]
         )
         self.proj_out = nn.Linear(inner_dim, in_channels)
+        self.video_length = 16
+
+    def set_video_length(self, video_length: int):
+        self.video_length = video_length
 
     def forward(self, hidden_states, encoder_hidden_states=None, attention_mask=None):
-        video_length = hidden_states.shape[0] // 2 # TODO: config this value in scripts
-
         batch, channel, height, weight = hidden_states.shape
         residual = hidden_states
 
@@ -159,7 +230,7 @@ class TemporalTransformer3DModel(nn.Module):
             hidden_states = block(
                 hidden_states,
                 encoder_hidden_states=encoder_hidden_states,
-                video_length=video_length,
+                video_length=self.video_length,
             )
 
         # output
@@ -204,15 +275,15 @@ class TemporalTransformerBlock(nn.Module):
             attention_blocks.append(
                 VersatileAttention(
                     attention_mode=block_name.split("_")[0],
-                    cross_attention_dim=cross_attention_dim
+                    context_dim=cross_attention_dim
                     if block_name.endswith("_Cross")
                     else None,
                     query_dim=dim,
                     heads=num_attention_heads,
                     dim_head=attention_head_dim,
                     dropout=dropout,
-                    bias=attention_bias,
+                    # bias=attention_bias, # remove for Comfy CrossAttention
-                    upcast_attention=upcast_attention,
+                    # upcast_attention=upcast_attention, # remove for Comfy CrossAttention
                     cross_frame_attention_mode=cross_frame_attention_mode,
                     temporal_position_encoding=temporal_position_encoding,
                     temporal_position_encoding_max_len=temporal_position_encoding_max_len,
@@ -284,7 +355,7 @@ class VersatileAttention(CrossAttention):
         assert attention_mode == "Temporal"
 
         self.attention_mode = attention_mode
-        self.is_cross_attention = kwargs["cross_attention_dim"] is not None
+        self.is_cross_attention = kwargs["context_dim"] is not None
 
         self.pos_encoder = (
             PositionalEncoding(
@@ -327,8 +398,8 @@ class VersatileAttention(CrossAttention):
         hidden_states = super().forward(
             hidden_states,
             encoder_hidden_states,
-            attention_mask,
+            value=None,
-            **cross_attention_kwargs,
+            mask=attention_mask,
         )
 
         hidden_states = rearrange(hidden_states, "(b d) f c -> (b f) d c", d=d)

animatediff/nodes.py

@@ -1,9 +1,10 @@
 import os
 import json
-import hashlib
 import torch
 import numpy as np
-from typing import Dict, List, Tuple
+from typing import Dict, List
+from torch import Tensor
+from torch.nn.functional import group_norm
 from PIL import Image
 from PIL.PngImagePlugin import PngInfo
 from einops import rearrange
@@ -11,19 +12,14 @@ 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.ldm.modules.diffusionmodules.util import GroupNorm32
 from comfy.utils import load_torch_file, calculate_parameters
-from comfy.model_patcher import ModelPatcher
 from nodes import KSampler
 
 from .logger import logger
 from .motion_module import MotionWrapper, VanillaTemporalModule
-from .model_utils import get_available_models, get_model_path, validate_mm_model
+from .model_utils import get_available_models, get_model_path, get_model_hash
-
-
-orig_forward_timestep_embed = openaimodel.forward_timestep_embed
-groupnorm32_original_forward = GroupNorm32.forward
 
 
 def forward_timestep_embed(
@@ -44,44 +40,43 @@ def forward_timestep_embed(
     return x
 
 
-def groupnorm32_mm_forward(self, x):
+def groupnorm_mm_factory(video_length: int):
-    x = rearrange(x, "(b f) c h w -> b c f h w", b=2)
+    def groupnorm_mm_forward(self, input: Tensor) -> Tensor:
-    x = groupnorm32_original_forward(self, x)
+        # axes_factor normalizes batch based on total conds and unconds passed in batch;
-    x = rearrange(x, "b c f h w -> (b f) c h w", b=2)
+        # the conds and unconds per batch can change based on VRAM optimizations that may kick in
-    return x
+        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] = {}
-original_model_hashs = set()
-injected_model_hashs: Dict[str, Tuple[str, str]] = {}
-
-
-def calculate_model_hash(unet):
-    t = unet.input_blocks[1]
-    m = hashlib.sha256()
-    for buf in t.buffers():
-        m.update(buf.cpu().numpy().view(np.uint8))
-    return m.hexdigest()
 
 
 def load_motion_module(model_name: str):
     model_path = get_model_path(model_name)
-    model_hash, is_v2 = validate_mm_model(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(model_name, is_v2=is_v2)
+        motion_module = MotionWrapper.from_pretrained(mm_state_dict, model_name)
 
-        parameters = calculate_parameters(mm_state_dict, "")
+        params = calculate_parameters(mm_state_dict, "")
-        usefp16 = model_management.should_use_fp16(model_params=parameters)
+        if model_management.should_use_fp16(model_params=params):
-        if usefp16:
+            logger.info(f"Converting motion module to fp16.")
-            logger.info("Using fp16, converting motion module to fp16")
             motion_module.half()
-        # offload_device = model_management.unet_offload_device()
+        offload_device = model_management.unet_offload_device()
-        # motion_module = motion_module.to(offload_device)
+        motion_module = motion_module.to(offload_device)
-        motion_module.load_state_dict(mm_state_dict)
+
         motion_modules[model_hash] = motion_module
 
     return motion_modules[model_hash]
@@ -176,81 +171,6 @@ ejectors = {
 }
 
 
-class AnimateDiffLoaderLegacy:
-    def __init__(self) -> None:
-        self.version = "legacy"
-
-    @classmethod
-    def INPUT_TYPES(s):
-        return {
-            "required": {
-                "model": ("MODEL",),
-                "model_name": (get_available_models(),),
-                "width": ("INT", {"default": 512, "min": 64, "max": 1024, "step": 8}),
-                "height": ("INT", {"default": 512, "min": 64, "max": 1024, "step": 8}),
-                "frame_number": (
-                    "INT",
-                    {"default": 16, "min": 2, "max": 24, "step": 1},
-                ),
-            },
-            "optional": {
-                "init_latent": ("LATENT",),
-            },
-        }
-
-    @classmethod
-    def IS_CHANGED(s, model: ModelPatcher):
-        unet = model.model.diffusion_model
-        # return calculate_model_hash(unet) not in injected_model_hashs
-        return hasattr(unet, "motion_module") and unet.motion_module is not None
-
-    RETURN_TYPES = ("MODEL", "LATENT")
-    CATEGORY = "Animate Diff"
-    FUNCTION = "inject_motion_modules"
-
-    def inject_motion_modules(
-        self,
-        model: ModelPatcher,
-        model_name: str,
-        width: int,
-        height: int,
-        frame_number=16,
-        init_latent: Dict[str, torch.Tensor] = None,
-    ):
-        motion_module = load_motion_module(model_name)
-
-        model = model.clone()
-        unet = model.model.diffusion_model
-        unet_hash = calculate_model_hash(unet)
-        need_inject = unet_hash not in injected_model_hashs
-
-        if unet_hash in injected_model_hashs:
-            (mm_hash, version) = injected_model_hashs[unet_hash]
-            if version != self.version or mm_hash != motion_module.mm_hash:
-                # injected by another motion module, unload first
-                logger.info(f"Ejecting motion module {mm_hash} version {version}.")
-                ejectors[version](unet)
-                need_inject = True
-            else:
-                logger.info(f"Motion module already injected, skipping injection.")
-
-        if need_inject:
-            logger.info(f"Injecting motion module {model_name} version {self.version}.")
-            injectors[self.version](unet, motion_module)
-            unet_hash = calculate_model_hash(unet)
-            injected_model_hashs[unet_hash] = (motion_module.mm_hash, self.version)
-
-        if init_latent is None:
-            latent = torch.zeros([frame_number, 4, height // 8, width // 8]).cpu()
-        else:
-            # clone value of first frame
-            latent = init_latent["samples"][:1, :, :, :].clone().cpu()
-            # repeat for all frames
-            latent = latent.repeat(frame_number, 1, 1, 1)
-
-        return (model, {"samples": latent})
-
-
 class AnimateDiffModuleLoader:
     @classmethod
     def INPUT_TYPES(s):
@@ -273,105 +193,6 @@ class AnimateDiffModuleLoader:
         return (motion_module,)
 
 
-class AnimateDiffLoader:
-    def __init__(self) -> None:
-        self.version = "v1"
-
-    @classmethod
-    def INPUT_TYPES(s):
-        return {
-            "required": {
-                "model": ("MODEL",),
-                "init_latent": ("LATENT",),
-                "model_name": (get_available_models(),),
-                "frame_number": (
-                    "INT",
-                    {"default": 16, "min": 2, "max": 32, "step": 1},
-                ),
-            },
-        }
-
-    @classmethod
-    def IS_CHANGED(s, model: ModelPatcher, _):
-        unet = model.model.diffusion_model
-        # return calculate_model_hash(unet) not in injected_model_hashs
-        return hasattr(unet, "motion_module") and unet.motion_module is not None
-
-    RETURN_TYPES = ("MODEL", "LATENT")
-    CATEGORY = "Animate Diff"
-    FUNCTION = "inject_motion_modules"
-
-    def inject_motion_modules(
-        self,
-        model: ModelPatcher,
-        init_latent: Dict[str, torch.Tensor],
-        model_name: str,
-        frame_number=16,
-    ):
-        motion_module = load_motion_module(model_name)
-
-        model = model.clone()
-        unet = model.model.diffusion_model
-        unet_hash = calculate_model_hash(unet)
-        need_inject = unet_hash not in injected_model_hashs
-
-        if unet_hash in injected_model_hashs:
-            (mm_type, version) = injected_model_hashs[unet_hash]
-            if version != self.version or mm_type != motion_module.mm_hash:
-                # injected by another motion module, unload first
-                logger.info(f"Ejecting motion module {mm_type} version {version}.")
-                ejectors[version](unet)
-                need_inject = True
-            else:
-                logger.info(f"Motion module already injected, skipping injection.")
-
-        if need_inject:
-            logger.info(f"Injecting motion module {model_name} version {self.version}.")
-            injectors[self.version](unet, motion_module)
-            unet_hash = calculate_model_hash(unet)
-            injected_model_hashs[unet_hash] = (motion_module.mm_hash, self.version)
-
-        init_frames = len(init_latent["samples"])
-        samples = init_latent["samples"][:init_frames, :, :, :].clone().cpu()
-
-        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)
-
-        return (model, {"samples": samples})
-
-
-class AnimateDiffUnload:
-    @classmethod
-    def INPUT_TYPES(s):
-        return {"required": {"model": ("MODEL",)}}
-
-    @classmethod
-    def IS_CHANGED(s, model: ModelPatcher):
-        unet = model.model.diffusion_model
-        return calculate_model_hash(unet) in injected_model_hashs
-
-    RETURN_TYPES = ("MODEL",)
-    CATEGORY = "Animate Diff"
-    FUNCTION = "unload_motion_modules"
-
-    def unload_motion_modules(self, model: ModelPatcher):
-        model = model.clone()
-        unet = model.model.diffusion_model
-        model_hash = calculate_model_hash(unet)
-        if model_hash in injected_model_hashs:
-            (model_name, version) = injected_model_hashs[model_hash]
-            logger.info(f"Ejecting motion module {model_name} version {version}.")
-            ejectors[version](unet)
-        else:
-            logger.info(f"Motion module not injected, skip unloading.")
-
-        return (model,)
-
-
 class AnimateDiffSampler(KSampler):
     @classmethod
     def INPUT_TYPES(s):
@@ -394,66 +215,57 @@ class AnimateDiffSampler(KSampler):
     def __init__(self) -> None:
         super().__init__()
         self.prev_beta = None
-        self.prev_alpha_cumprod = None
+        self.prev_linear_start = None
-        self.prev_alpha_cumprod_prev = None
+        self.prev_linear_end = None
 
-    def override_ddim_alpha(self, model):
+    def override_beta_schedule(self, model: BaseModel):
-        logger.info(f"Setting DDIM alpha.")
+        logger.info(f"Override beta schedule.")
-        device = model_management.unet_offload_device()
+        self.prev_beta = model.get_buffer("betas").cpu().clone()
-
+        self.prev_linear_start = model.linear_start
-        beta_start = 0.00085
+        self.prev_linear_end = model.linear_end
-        beta_end = 0.012
+        model.register_schedule(
-        betas = torch.linspace(
+            given_betas=None,
-            beta_start,
+            beta_schedule="sqrt_linear",
-            beta_end,
+            timesteps=1000,
-            model.num_timesteps,
+            linear_start=0.00085,
-            dtype=torch.float32,
+            linear_end=0.012,
-            device=device,
+            cosine_s=8e-3,
         )
-        alphas = 1.0 - betas
-        alphas_cumprod = torch.cumprod(alphas, dim=0)
-        alphas_cumprod_prev = torch.cat(
-            (
-                torch.tensor([1.0], dtype=torch.float32, device=device),
-                alphas_cumprod[:-1],
-            )
-        )
-        self.prev_beta = model.betas
-        model.betas = betas
-        self.prev_alpha_cumprod = model.alphas_cumprod
-        model.alphas_cumprod = alphas_cumprod
-        self.prev_alpha_cumprod_prev = model.alphas_cumprod_prev
-        model.alphas_cumprod_prev = alphas_cumprod_prev
 
-    def restore_ddim_alpha(self, model):
+    def restore_beta_schedule(self, model: BaseModel):
-        logger.info(f"Restoring DDIM alpha.")
+        logger.info(f"Restoring beta schedule.")
-        model.betas = self.prev_beta
+        model.register_schedule(
-        model.alphas_cumprod = self.prev_alpha_cumprod
+            given_betas=self.prev_beta,
-        model.alphas_cumprod_prev = self.prev_alpha_cumprod_prev
+            linear_start=self.prev_linear_start,
+            linear_end=self.prev_linear_end,
+        )
         self.prev_beta = None
-        self.prev_alpha_cumprod = None
+        self.prev_linear_start = None
-        self.prev_alpha_cumprod_prev = None
+        self.prev_linear_end = None
 
-    def inject_motion_module(self, model, motion_module, inject_method):
+    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_ddim_alpha(model.model)
+        self.override_beta_schedule(model.model)
         if not motion_module.is_v2:
-            logger.info(f"Hacking GroupNorm32 forward function.")
+            logger.info(f"Hacking GroupNorm.forward function.")
-            GroupNorm32.forward = groupnorm32_mm_forward
+            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_ddim_alpha(model.model)
+        self.restore_beta_schedule(model.model)
         if not unet.motion_module.is_v2:
-            logger.info(f"Restore GroupNorm32 forward function.")
+            logger.info(f"Restore GroupNorm.forward function.")
-            GroupNorm32.forward = groupnorm32_original_forward
+            torch.nn.GroupNorm.forward = orig_groupnorm_forward
 
         logger.info(f"Ejecting motion module with method {inject_method}.")
         ejectors[inject_method](unet)
@@ -474,7 +286,9 @@ class AnimateDiffSampler(KSampler):
         latent_image,
         denoise=1.0,
     ):
-        model = self.inject_motion_module(model, motion_module, inject_method)
+        model = self.inject_motion_module(
+            model, motion_module, inject_method, frame_number
+        )
 
         init_frames = len(latent_image["samples"])
         samples = latent_image["samples"][:init_frames, :, :, :].clone().cpu()
@@ -488,22 +302,23 @@ class AnimateDiffSampler(KSampler):
 
         latent_image = {"samples": samples}
 
-        results = super().sample(
+        try:
-            model,
+            return super().sample(
-            seed,
+                model,
-            steps,
+                seed,
-            cfg,
+                steps,
-            sampler_name,
+                cfg,
-            scheduler,
+                sampler_name,
-            positive,
+                scheduler,
-            negative,
+                positive,
-            latent_image,
+                negative,
-            denoise=1.0,
+                latent_image,
-        )
+                denoise=denoise,
-
+            )
-        self.eject_motion_module(model, inject_method)
+        except:
-
+            raise
-        return results
+        finally:
+            self.eject_motion_module(model, inject_method)
 
 
 class AnimateDiffCombine:
@@ -517,8 +332,11 @@ class AnimateDiffCombine:
                     {"default": 8, "min": 1, "max": 24, "step": 1},
                 ),
                 "loop_count": ("INT", {"default": 0, "min": 0, "max": 100, "step": 1}),
-                "save_image": (["Enabled", "Disabled"],),
+                "save_image": ([True, False],),
-                "filename_prefix": ("STRING", {"default": "AnimateDiff"}),
+                "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],),
             },
             "hidden": {
                 "prompt": "PROMPT",
@@ -526,7 +344,7 @@ class AnimateDiffCombine:
             },
         }
 
-    RETURN_TYPES = ()
+    RETURN_TYPES = ("GIF",)
     OUTPUT_NODE = True
     CATEGORY = "Animate Diff"
     FUNCTION = "generate_gif"
@@ -536,22 +354,24 @@ class AnimateDiffCombine:
         images,
         frame_rate: int,
         loop_count: int,
-        save_image="Enabled",
+        save_image=True,
         filename_prefix="AnimateDiff",
+        format="image/gif",
+        pingpong=False,
         prompt=None,
         extra_pnginfo=None,
     ):
         # convert images to numpy
-        pil_images: List[Image.Image] = []
+        frames: List[Image.Image] = []
         for image in images:
             img = 255.0 * image.cpu().numpy()
             img = Image.fromarray(np.clip(img, 0, 255).astype(np.uint8))
-            pil_images.append(img)
+            frames.append(img)
 
         # save image
         output_dir = (
             folder_paths.get_output_directory()
-            if save_image == "Enabled"
+            if save_image
             else folder_paths.get_temp_directory()
         )
         (
@@ -572,48 +392,70 @@ class AnimateDiffCombine:
         # save first frame as png to keep metadata
         file = f"{filename}_{counter:05}_.png"
         file_path = os.path.join(full_output_folder, file)
-        pil_images[0].save(
+        frames[0].save(
             file_path,
             pnginfo=metadata,
             compress_level=4,
         )
+        if pingpong:
+            frames = frames + frames[-2:0:-1]
 
-        # save gif
+        format_type, format_ext = format.split("/")
-        file = f"{filename}_{counter:05}_.gif"
-        file_path = os.path.join(full_output_folder, file)
-        pil_images[0].save(
-            file_path,
-            save_all=True,
-            append_images=pil_images[1:],
-            duration=round(1000 / frame_rate),
-            loop=loop_count,
-            compress_level=4,
-        )
 
-        print("Saved gif to", file_path, os.path.exists(file_path))
+        if format_type == "image":
+            file = f"{filename}_{counter:05}_.{format_ext}"
+            file_path = os.path.join(full_output_folder, file)
+            frames[0].save(
+                file_path,
+                format=format_ext.upper(),
+                save_all=True,
+                append_images=frames[1:],
+                duration=round(1000 / frame_rate),
+                loop=loop_count,
+                compress_level=4,
+            )
+        else:
+            # save webm
+            import shutil
+            import subprocess
+
+            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 = 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]
+
+            env=os.environ
+            if "environment" in video_format:
+                env.update(video_format["environment"])
+            with subprocess.Popen(args, stdin=subprocess.PIPE, env=env) as proc:
+                for frame in frames:
+                    proc.stdin.write(frame.tobytes())
 
         previews = [
             {
                 "filename": file,
                 "subfolder": subfolder,
-                "type": "output" if save_image == "Enabled" else "temp",
+                "type": "output" if save_image else "temp",
+                "format": format,
             }
         ]
-        return {"ui": {"images": previews}}
+        return {"ui": {"gifs": previews}}
 
 
 NODE_CLASS_MAPPINGS = {
-    # "AnimateDiffLoader": AnimateDiffLoaderLegacy,
-    # "AnimateDiffLoader_v2": AnimateDiffLoader,
-    # "AnimateDiffUnload": AnimateDiffUnload,
     "AnimateDiffModuleLoader": AnimateDiffModuleLoader,
     "AnimateDiffCombine": AnimateDiffCombine,
     "AnimateDiffSampler": AnimateDiffSampler,
 }
 NODE_DISPLAY_NAME_MAPPINGS = {
-    # "AnimateDiffLoader": "[DEPRECATED] Animate Diff Loader Legacy",
-    # "AnimateDiffLoader_v2": "[DEPRECATED] Animate Diff Loader",
-    # "AnimateDiffUnload": "[DEPRECATED] Animate Diff Unload",
     "AnimateDiffModuleLoader": "Animate Diff Module Loader",
     "AnimateDiffSampler": "Animate Diff Sampler",
     "AnimateDiffCombine": "Animate Diff Combine",

video_formats/av1-webm.json

@@ -0,0 +1,10 @@
+{
+    "main_pass":
+    [
+        "-n", "-c:v", "libsvtav1",
+        "-pix_fmt", "yuv420p10le",
+        "-crf", "23"
+    ],
+     "extension": "webm",
+     "environment": {"SVT_LOG": "1"}
+}

video_formats/h264-mp4.json

@@ -0,0 +1,9 @@
+{
+    "main_pass":
+    [
+        "-n", "-c:v", "libx264",
+        "-pix_fmt", "yuv420p",
+        "-crf", "19"
+    ],
+     "extension": "mp4"
+}

video_formats/h265-mp4.json

@@ -0,0 +1,11 @@
+{
+    "main_pass":
+    [
+        "-n", "-c:v", "libx265",
+        "-pix_fmt", "yuv420p10le",
+        "-preset", "medium",
+        "-crf", "22",
+        "-x265-params", "log-level=quiet"
+    ],
+     "extension": "mp4"
+}

video_formats/webm.json

@@ -0,0 +1,9 @@
+{
+    "main_pass":
+    [
+        "-n",
+        "-pix_fmt", "yuv420p",
+        "-crf", "23"
+    ],
+     "extension": "webm"
+}

web/js/gif_preview.js

@@ -0,0 +1,146 @@
+import { app } from '../../../scripts/app.js'
+import { api } from '../../../scripts/api.js'
+
+function offsetDOMWidget(
+  widget,
+  ctx,
+  node,
+  widgetWidth,
+  widgetY,
+  height
+) {
+  const margin = 10
+  const elRect = ctx.canvas.getBoundingClientRect()
+  const transform = new DOMMatrix()
+    .scaleSelf(
+      elRect.width / ctx.canvas.width,
+      elRect.height / ctx.canvas.height
+    )
+    .multiplySelf(ctx.getTransform())
+    .translateSelf(0, widgetY + margin)
+
+  const scale = new DOMMatrix().scaleSelf(transform.a, transform.d)
+  Object.assign(widget.inputEl.style, {
+    transformOrigin: '0 0',
+    transform: scale,
+    left: `${transform.e}px`,
+    top: `${transform.d + transform.f}px`,
+    width: `${widgetWidth}px`,
+    height: `${(height || widget.parent?.inputHeight || 32) - margin}px`,
+    position: 'absolute',
+    background: !node.color ? '' : node.color,
+    color: !node.color ? '' : 'white',
+    zIndex: 5, //app.graph._nodes.indexOf(node),
+  })
+}
+
+export const hasWidgets = (node) => {
+  if (!node.widgets || !node.widgets?.[Symbol.iterator]) {
+    return false
+  }
+  return true
+}
+
+export const cleanupNode = (node) => {
+  if (!hasWidgets(node)) {
+    return
+  }
+
+  for (const w of node.widgets) {
+    if (w.canvas) {
+      w.canvas.remove()
+    }
+    if (w.inputEl) {
+      w.inputEl.remove()
+    }
+    // calls the widget remove callback
+    w.onRemoved?.()
+  }
+}
+
+const CreatePreviewElement = (name, val, format) => {
+  const [type] = format.split('/')
+
+  const w = {
+    name,
+    type,
+    value: val,
+    draw: function (ctx, node, widgetWidth, widgetY, height) {
+      const [cw, ch] = this.computeSize(widgetWidth)
+      offsetDOMWidget(this, ctx, node, widgetWidth, widgetY, ch)
+    },
+    computeSize: function (_) {
+      const ratio = this.inputRatio || 1
+      const width = Math.max(220, this.parent.size[0])
+      return [width, (width / ratio + 10)]
+    },
+    onRemoved: function () {
+      if (this.inputEl) {
+        this.inputEl.remove()
+      }
+    },
+  }
+
+  w.inputEl = document.createElement(type === 'video' ? 'video' : 'img')
+  w.inputEl.src = w.value
+  if (type === 'video') {
+    w.inputEl.setAttribute('type', 'video/webm');
+    w.inputEl.autoplay = true
+    w.inputEl.loop = true
+    w.inputEl.controls = false;
+  }
+  w.inputEl.onload = function () {
+    w.inputRatio = w.inputEl.naturalWidth / w.inputEl.naturalHeight
+  }
+  document.body.appendChild(w.inputEl)
+  return w
+}
+
+const gif_preview = {
+  name: 'AnimateDiff.gif_preview',
+  async beforeRegisterNodeDef(nodeType, nodeData, app) {
+    switch (nodeData.name) {
+      case 'AnimateDiffCombine': {
+        const onExecuted = nodeType.prototype.onExecuted
+        nodeType.prototype.onExecuted = function (message) {
+          const prefix = 'ad_gif_preview_'
+          const r = onExecuted ? onExecuted.apply(this, message) : undefined
+
+          if (this.widgets) {
+            const pos = this.widgets.findIndex((w) => w.name === `${prefix}_0`)
+            if (pos !== -1) {
+              for (let i = pos; i < this.widgets.length; i++) {
+                this.widgets[i].onRemoved?.()
+              }
+              this.widgets.length = pos
+            }
+            if (message?.gifs) {
+              message.gifs.forEach((params, i) => {
+                const previewUrl = api.apiURL(
+                  '/view?' + new URLSearchParams(params).toString()
+                )
+                const w = this.addCustomWidget(
+                  CreatePreviewElement(`${prefix}_${i}`, previewUrl, params.format || 'image/gif')
+                )
+                w.parent = this
+              })
+            }
+
+            const onRemoved = this.onRemoved
+            this.onRemoved = () => {
+              cleanupNode(this)
+              return onRemoved?.()
+            }
+          }
+
+          // keep width and update height
+          this.setSize([this.size[0], this.computeSize([this.size[0], this.size[1]])[1]])
+          return r
+        }
+        break
+      }
+    }
+  }
+}
+
+app.registerExtension(gif_preview)

workflow.json

@@ -0,0 +1,451 @@
+{
+  "last_node_id": 20,
+  "last_link_id": 35,
+  "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": [
+        "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": 7,
+      "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": 8,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "link": 19
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "AnimateDiffCombine"
+      },
+      "widgets_values": [
+        8,
+        0,
+        "Enabled",
+        "AnimateDiff"
+      ]
+    },
+    {
+      "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": 15,
+      "type": "AnimateDiffSampler",
+      "pos": [
+        882,
+        192
+      ],
+      "size": {
+        "0": 315,
+        "1": 330
+      },
+      "flags": {},
+      "order": 6,
+      "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
+        }
+      ],
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": [
+            28
+          ],
+          "shape": 3,
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "AnimateDiffSampler"
+      },
+      "widgets_values": [
+        "default",
+        16,
+        345029849956677,
+        "fixed",
+        20,
+        8,
+        "euler",
+        "normal",
+        0.8
+      ]
+    },
+    {
+      "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": [
+        "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
+      ]
+    }
+  ],
+  "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"
+    ]
+  ],
+  "groups": [],
+  "config": {},
+  "extra": {},
+  "version": 0.4
+}