nodes/cubiq-ComfyUI_IPAdapter_plus
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

preliminary support for FaceID

Browse Source
This commit is contained in:
matt3o
2023-12-22 15:47:45 +01:00
parent a4e71ccbbf
commit 69e82da76b
5 changed files with 1750 additions and 63 deletions
Download Patch File Download Diff File Expand all files Collapse all files
IPAdapterPlus.py
+191 -61
View File
@@ -25,6 +25,8 @@ else:
current_paths, _ = folder_paths.folder_names_and_paths["ipadapter"]
folder_paths.folder_names_and_paths["ipadapter"] = (current_paths, folder_paths.supported_pt_extensions)
INSIGHTFACE_DIR = os.path.join(folder_paths.models_dir, "insightface")
class MLPProjModel(torch.nn.Module):
"""SD model with image prompt"""
def __init__(self, cross_attention_dim=1024, clip_embeddings_dim=1024):
@@ -41,6 +43,27 @@ class MLPProjModel(torch.nn.Module):
clip_extra_context_tokens = self.proj(image_embeds)
return clip_extra_context_tokens
class MLPProjModelFaceId(torch.nn.Module):
"""SD model with image prompt"""
def __init__(self, cross_attention_dim=768, id_embeddings_dim=512, num_tokens=4):
super().__init__()
self.cross_attention_dim = cross_attention_dim
self.num_tokens = num_tokens
self.proj = torch.nn.Sequential(
torch.nn.Linear(id_embeddings_dim, id_embeddings_dim*2),
torch.nn.GELU(),
torch.nn.Linear(id_embeddings_dim*2, cross_attention_dim*num_tokens),
)
self.norm = torch.nn.LayerNorm(cross_attention_dim)
def forward(self, id_embeds):
clip_extra_context_tokens = self.proj(id_embeds)
clip_extra_context_tokens = clip_extra_context_tokens.reshape(-1, self.num_tokens, self.cross_attention_dim)
clip_extra_context_tokens = self.norm(clip_extra_context_tokens)
return clip_extra_context_tokens
class ImageProjModel(nn.Module):
def __init__(self, cross_attention_dim=1024, clip_embeddings_dim=1024, clip_extra_context_tokens=4):
super().__init__()
@@ -153,8 +176,24 @@ def contrast_adaptive_sharpening(image, amount):
return (output)
def tensorToCV(image):
import numpy as np
# TODO: there must be a better way
out = []
image = image.detach().cpu()
for i in range(image.shape[0]):
img = image[i]
img = img[..., [2, 1, 0]] # Convert from RGB to BGR
img = torch.clamp(img * 255, 0, 255).to(torch.uint8) # Scale pixel values up to [0, 255] if they are in [0, 1]
img = img.numpy()
out.append(img)
out = np.stack(img, axis=0)
return out
class IPAdapter(nn.Module):
def __init__(self, ipadapter_model, cross_attention_dim=1024, output_cross_attention_dim=1024, clip_embeddings_dim=1024, clip_extra_context_tokens=4, is_sdxl=False, is_plus=False, is_full=False):
def __init__(self, ipadapter_model, cross_attention_dim=1024, output_cross_attention_dim=1024, clip_embeddings_dim=1024, clip_extra_context_tokens=4, is_sdxl=False, is_plus=False, is_full=False, is_faceid=False):
super().__init__()
self.clip_embeddings_dim = clip_embeddings_dim
@@ -164,7 +203,13 @@ class IPAdapter(nn.Module):
self.is_sdxl = is_sdxl
self.is_full = is_full
self.image_proj_model = self.init_proj() if not is_plus else self.init_proj_plus()
if is_faceid:
self.image_proj_model = self.init_proj_faceid()
elif is_plus:
self.image_proj_model = self.init_proj_plus()
else:
self.image_proj_model = self.init_proj()
self.image_proj_model.load_state_dict(ipadapter_model["image_proj"])
self.ip_layers = To_KV(ipadapter_model["ip_adapter"])
@@ -195,6 +240,14 @@ class IPAdapter(nn.Module):
)
return image_proj_model
def init_proj_faceid(self):
image_proj_model = MLPProjModelFaceId(
cross_attention_dim=self.cross_attention_dim,
id_embeddings_dim=512,
num_tokens=self.clip_extra_context_tokens,
)
return image_proj_model
@torch.inference_mode()
def get_image_embeds(self, clip_embed, clip_embed_zeroed):
image_prompt_embeds = self.image_proj_model(clip_embed)
@@ -385,6 +438,31 @@ class IPAdapterModelLoader:
return (model,)
class InsightFaceLoader:
@classmethod
def INPUT_TYPES(s):
return {
"required": {
"provider": (["CPU", "CUDA", "ROCM"], ),
},
}
RETURN_TYPES = ("CLIP_VISION",)
FUNCTION = "load_insight_face"
CATEGORY = "ipadapter"
def load_insight_face(self, provider):
try:
from insightface.app import FaceAnalysis
except ImportError:
raise Exception('IPAdapter: InsightFace is not installed! Install the missing dependencies if you wish to use FaceID models.')
model = FaceAnalysis(name="buffalo_l", root=INSIGHTFACE_DIR, providers=[provider + 'ExecutionProvider',])
model.prepare(ctx_id=0, det_size=(640, 640))
return (model,)
class IPAdapterApply:
@classmethod
def INPUT_TYPES(s):
@@ -415,7 +493,8 @@ class IPAdapterApply:
self.device = comfy.model_management.get_torch_device()
self.weight = weight
self.is_full = "proj.0.weight" in ipadapter["image_proj"]
self.is_plus = self.is_full or "latents" in ipadapter["image_proj"]
self.is_faceid = "0.to_q_lora.down.weight" in ipadapter["ip_adapter"] # TODO: better way to detect faceid?
self.is_plus = (self.is_full or "latents" in ipadapter["image_proj"]) and not self.is_faceid
output_cross_attention_dim = ipadapter["ip_adapter"]["1.to_k_ip.weight"].shape[1]
self.is_sdxl = output_cross_attention_dim == 2048
@@ -427,24 +506,33 @@ class IPAdapterApply:
clip_embed = embeds[0].cpu()
clip_embed_zeroed = embeds[1].cpu()
else:
if image.shape[1] != image.shape[2]:
print("\033[33mINFO: the IPAdapter reference image is not a square, CLIPImageProcessor will resize and crop it at the center. If the main focus of the picture is not in the middle the result might not be what you are expecting.\033[0m")
if self.is_faceid:
clip_embed = clip_vision.get(tensorToCV(image)) # TODO: support multiple images (is it needed?)
if not clip_embed:
print("\033[33mWARNING!!! InsightFace wasn't able to detect the face. Try to use the PrepImageForInsightFace node.\033[0m")
clip_embed = clip_vision.encode_image(image)
neg_image = image_add_noise(image, noise) if noise > 0 else None
if self.is_plus:
clip_embed = clip_embed.penultimate_hidden_states
if noise > 0:
clip_embed_zeroed = clip_vision.encode_image(neg_image).penultimate_hidden_states
else:
clip_embed_zeroed = zeroed_hidden_states(clip_vision, image.shape[0])
#clip_embed = torch.from_numpy(clip_embed[0].normed_embedding).unsqueeze(0)
clip_embed = torch.stack([torch.from_numpy(elem.normed_embedding).unsqueeze(0) for elem in clip_embed], dim=0)
clip_embed_zeroed = torch.zeros_like(clip_embed)
else:
clip_embed = clip_embed.image_embeds
if noise > 0:
clip_embed_zeroed = clip_vision.encode_image(neg_image).image_embeds
if image.shape[1] != image.shape[2]:
print("\033[33mINFO: the IPAdapter reference image is not a square, CLIPImageProcessor will resize and crop it at the center. If the main focus of the picture is not in the middle the result might not be what you are expecting.\033[0m")
clip_embed = clip_vision.encode_image(image)
neg_image = image_add_noise(image, noise) if noise > 0 else None
if self.is_plus:
clip_embed = clip_embed.penultimate_hidden_states
if noise > 0:
clip_embed_zeroed = clip_vision.encode_image(neg_image).penultimate_hidden_states
else:
clip_embed_zeroed = zeroed_hidden_states(clip_vision, image.shape[0])
else:
clip_embed_zeroed = torch.zeros_like(clip_embed)
clip_embed = clip_embed.image_embeds
if noise > 0:
clip_embed_zeroed = clip_vision.encode_image(neg_image).image_embeds
else:
clip_embed_zeroed = torch.zeros_like(clip_embed)
clip_embeddings_dim = clip_embed.shape[-1]
@@ -457,6 +545,7 @@ class IPAdapterApply:
is_sdxl=self.is_sdxl,
is_plus=self.is_plus,
is_full=self.is_full,
is_faceid=self.is_faceid,
)
self.ipadapter.to(self.device, dtype=self.dtype)
@@ -511,6 +600,83 @@ class IPAdapterApply:
return (work_model, )
def prepImage(image, interpolation="LANCZOS", crop_position="center", size=(224,224), sharpening=0.0, padding=0):
_, oh, ow, _ = image.shape
output = image.permute([0,3,1,2])
if "pad" in crop_position:
target_length = max(oh, ow)
pad_l = (target_length - ow) // 2
pad_r = (target_length - ow) - pad_l
pad_t = (target_length - oh) // 2
pad_b = (target_length - oh) - pad_t
output = F.pad(output, (pad_l, pad_r, pad_t, pad_b), value=0, mode="constant")
else:
crop_size = min(oh, ow)
x = (ow-crop_size) // 2
y = (oh-crop_size) // 2
if "top" in crop_position:
y = 0
elif "bottom" in crop_position:
y = oh-crop_size
elif "left" in crop_position:
x = 0
elif "right" in crop_position:
x = ow-crop_size
x2 = x+crop_size
y2 = y+crop_size
# crop
output = output[:, :, y:y2, x:x2]
# resize (apparently PIL resize is better than tourchvision interpolate)
imgs = []
for i in range(output.shape[0]):
img = TT.ToPILImage()(output[i])
img = img.resize(size, resample=Image.Resampling[interpolation])
imgs.append(TT.ToTensor()(img))
output = torch.stack(imgs, dim=0)
imgs = None # zelous GC
if sharpening > 0:
output = contrast_adaptive_sharpening(output, sharpening)
if padding > 0:
output = F.pad(output, (padding, padding, padding, padding), value=255, mode="constant")
output = output.permute([0,2,3,1])
return output
class PrepImageForInsightFace:
@classmethod
def INPUT_TYPES(s):
return {"required": {
"image": ("IMAGE",),
"crop_position": (["center", "top", "bottom", "left", "right"],),
"sharpening": ("FLOAT", {"default": 0.0, "min": 0, "max": 1, "step": 0.05}),
"pad_around": ("BOOLEAN", { "default": True }),
},
}
RETURN_TYPES = ("IMAGE",)
FUNCTION = "prep_image"
CATEGORY = "ipadapter"
def prep_image(self, image, crop_position, sharpening=0.0, pad_around=True):
if pad_around:
padding = 60
size = (580, 580)
else:
padding = 0
size = (640, 640)
output = prepImage(image, "LANCZOS", crop_position, size, sharpening, padding)
return (output, )
class PrepImageForClipVision:
@classmethod
def INPUT_TYPES(s):
@@ -528,49 +694,9 @@ class PrepImageForClipVision:
CATEGORY = "ipadapter"
def prep_image(self, image, interpolation="LANCZOS", crop_position="center", sharpening=0.0):
_, oh, ow, _ = image.shape
output = image.permute([0,3,1,2])
if "pad" in crop_position:
target_length = max(oh, ow)
pad_l = (target_length - ow) // 2
pad_r = (target_length - ow) - pad_l
pad_t = (target_length - oh) // 2
pad_b = (target_length - oh) - pad_t
output = F.pad(output, (pad_l, pad_r, pad_t, pad_b), value=0, mode="constant")
else:
crop_size = min(oh, ow)
x = (ow-crop_size) // 2
y = (oh-crop_size) // 2
if "top" in crop_position:
y = 0
elif "bottom" in crop_position:
y = oh-crop_size
elif "left" in crop_position:
x = 0
elif "right" in crop_position:
x = ow-crop_size
x2 = x+crop_size
y2 = y+crop_size
# crop
output = output[:, :, y:y2, x:x2]
# resize (apparently PIL resize is better than tourchvision interpolate)
imgs = []
for i in range(output.shape[0]):
img = TT.ToPILImage()(output[i])
img = img.resize((224,224), resample=Image.Resampling[interpolation])
imgs.append(TT.ToTensor()(img))
output = torch.stack(imgs, dim=0)
if sharpening > 0:
output = contrast_adaptive_sharpening(output, sharpening)
output = output.permute([0,2,3,1])
return (output,)
size = (224, 224)
output = prepImage(image, interpolation, crop_position, size, sharpening, 0)
return (output, )
class IPAdapterEncoder:
@classmethod
@@ -737,6 +863,8 @@ NODE_CLASS_MAPPINGS = {
"IPAdapterSaveEmbeds": IPAdapterSaveEmbeds,
"IPAdapterLoadEmbeds": IPAdapterLoadEmbeds,
"IPAdapterBatchEmbeds": IPAdapterBatchEmbeds,
"InsightFaceLoader": InsightFaceLoader,
"PrepImageForInsightFace": PrepImageForInsightFace,
}
NODE_DISPLAY_NAME_MAPPINGS = {
@@ -748,4 +876,6 @@ NODE_DISPLAY_NAME_MAPPINGS = {
"IPAdapterSaveEmbeds": "Save IPAdapter Embeds",
"IPAdapterLoadEmbeds": "Load IPAdapter Embeds",
"IPAdapterBatchEmbeds": "IPAdapter Batch Embeds",
"InsightFaceLoader": "Load InsightFace",
"PrepImageForInsightFace": "Prepare Image For InsightFace",
}
README.md
+28 -2
View File
@@ -5,6 +5,8 @@ IPAdapter implementation that follows the ComfyUI way of doing things. The code
## Important updates
**2023/12/22**: Added support for FaceID models. [Read the documentation](#faceid) for details.
**2023/12/05**: Added `batch embeds` node. This lets you encode images in batches and merge them together into an `IPAdapter Apply Encoded` node. Useful mostly for animations because the clip vision encoder takes a lot of VRAM. My suggestion is to split the animation in batches of about 120 frames.
**2023/11/29**: Added `unfold_batch` option to send the reference images sequentially to a latent batch. Useful for animations.
@@ -21,8 +23,6 @@ IPAdapter implementation that follows the ComfyUI way of doing things. The code
**2023/11/02**: Added compatibility with the new models in safetensors format (available on [huggingface](https://huggingface.co/h94/IP-Adapter)).
**2023/10/12**: Added image weighting in the `IPAdapterEncoder` node. This update is somewhat breaking; if you use `IPAdapterEncoder` and `PrepImageForClipVision` nodes you need to remove them from your workflow, refresh and recreate them. In the examples you'll find a [workflow](examples/IPAdapter_weighted.json) for weighted images.
*(previous updates removed for better readability)*
## What is it?
@@ -81,6 +81,12 @@ You can rename them to something easier to remember or put them into a sub-direc
**Note:** the image encoders are actually [ViT-H](https://huggingface.co/laion/CLIP-ViT-H-14-laion2B-s32B-b79K) and [ViT-bigG](https://huggingface.co/laion/CLIP-ViT-bigG-14-laion2B-39B-b160k) (used only for one SDXL model). You probably already have them.
For **FaceID** you need:
- The [main SD1.5 model](https://huggingface.co/h94/IP-Adapter-FaceID/resolve/main/ip-adapter-faceid_sd15.bin) to be placed into the ipadapter models directory.
- The [Lora](https://huggingface.co/h94/IP-Adapter-FaceID/resolve/main/ip-adapter-faceid_sd15_lora.safetensors) to be planced into `ComfyUI/models/loras/` directory.
**Note:** FaceID requires `insightface` and `onnxruntime` (you probably have the latter installer already) that you need to install inside your ComfyUI environment with `pip`. This is not a mandatory requirement, the extension will still work, just without FaceID support.
## How to
There's a basic workflow included in this repo and a few examples in the [examples](./examples/) directory. Usually it's a good idea to lower the `weight` to at least `0.8`.
@@ -185,6 +191,26 @@ In the `Apply IPAdapter` node you can set a start and an end point. The IPAdapte
<img src="./examples/timestepping.jpg" width="100%" alt="timestepping" />
### FaceID
FaceID is a new IPAdapter model that takes the embeddings from [InsightFace](https://github.com/deepinsight/insightface). As such you need to install `insightface` in your ComfyUI python environment. You may also need `onnxruntime` and `onnxruntime-gpu`. Note that your CUDA version might not be compatible with onnxruntime, in that case you can select the "CPU" provider from the `Load InsightFace model` node.
The first time you use InsightFace the model will be downloaded automatically, check the console to see the progress. If you get an error you need to donwload the [buffalo_l](https://github.com/deepinsight/insightface/releases) model manually inside the `ComfyUI/models/insightface/models` directory. Also every time you run the workflow for the first time InsightFace will take quite a few seconds to load.
**The FaceID model is used in conjuction with its Lora!** Check the [installation instructions](#installation) for the links to all models.
The reference image needs to be prepared differently compared to the other IPAdapter face models. While standard face models expects the face take basically the whole frame, FaceID prefers the subject to be a little further away. Don't cut the face too close and leave hair, beard, ears, neck in the picture.
**InsightFace will often fail to detect the face** and it will throw an error. Try with a different picture possibly cut to half-bust. FaceID generally works with drawings/illustrations too and the result is often very nice.
I just implemented the FaceID code so I don't have best practices yet and more testing is needed. It's important to understand that **FaceID can (and should) be used as a first pass for an additional IPAdapter Face model**.
In the [examples directory](./examples/) you'll find a few workflows to get you started with FaceID.
The following would be a basic workflow that includes FaceID enhanced by a Plus Face model.
<img src="./examples/face_id_wf.jpg" width="100%" alt="timestepping" />
## Troubleshooting
Please check the [troubleshooting](https://github.com/cubiq/ComfyUI_IPAdapter_plus/issues/108) before posting a new issue.
examples/FaceID_basic.json
+656
View File
@@ -0,0 +1,656 @@
{
"last_node_id": 30,
"last_link_id": 78,
"nodes": [
{
"id": 6,
"type": "VAEDecode",
"pos": [
1670,
620
],
"size": {
"0": 140,
"1": 50
},
"flags": {},
"order": 11,
"mode": 0,
"inputs": [
{
"name": "samples",
"type": "LATENT",
"link": 7
},
{
"name": "vae",
"type": "VAE",
"link": 8
}
],
"outputs": [
{
"name": "IMAGE",
"type": "IMAGE",
"links": [
72
],
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "VAEDecode"
}
},
{
"id": 7,
"type": "VAELoader",
"pos": [
1380,
920
],
"size": {
"0": 240,
"1": 60
},
"flags": {},
"order": 0,
"mode": 0,
"outputs": [
{
"name": "VAE",
"type": "VAE",
"links": [
8
],
"shape": 3,
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "VAELoader"
},
"widgets_values": [
"vae-ft-mse-840000-ema-pruned.safetensors"
]
},
{
"id": 12,
"type": "InsightFaceLoader",
"pos": [
710,
240
],
"size": {
"0": 210,
"1": 58
},
"flags": {},
"order": 1,
"mode": 0,
"outputs": [
{
"name": "CLIP_VISION",
"type": "CLIP_VISION",
"links": [
11
],
"shape": 3,
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "InsightFaceLoader"
},
"widgets_values": [
"CPU"
],
"color": "#223",
"bgcolor": "#335"
},
{
"id": 9,
"type": "IPAdapterModelLoader",
"pos": [
630,
130
],
"size": {
"0": 300,
"1": 60
},
"flags": {},
"order": 2,
"mode": 0,
"outputs": [
{
"name": "IPADAPTER",
"type": "IPADAPTER",
"links": [
10
],
"shape": 3,
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "IPAdapterModelLoader"
},
"widgets_values": [
"ip-adapter-faceid_sd15.bin"
],
"color": "#223",
"bgcolor": "#335"
},
{
"id": 2,
"type": "CheckpointLoaderSimple",
"pos": [
50,
700
],
"size": {
"0": 290,
"1": 100
},
"flags": {},
"order": 3,
"mode": 0,
"outputs": [
{
"name": "MODEL",
"type": "MODEL",
"links": [
15
],
"slot_index": 0
},
{
"name": "CLIP",
"type": "CLIP",
"links": [
5,
6
],
"slot_index": 1
},
{
"name": "VAE",
"type": "VAE",
"links": [],
"slot_index": 2
}
],
"properties": {
"Node name for S&R": "CheckpointLoaderSimple"
},
"widgets_values": [
"sd15/realisticVisionV51_v51VAE.safetensors"
]
},
{
"id": 4,
"type": "CLIPTextEncode",
"pos": [
710,
570
],
"size": {
"0": 400,
"1": 160
},
"flags": {},
"order": 7,
"mode": 0,
"inputs": [
{
"name": "clip",
"type": "CLIP",
"link": 5
}
],
"outputs": [
{
"name": "CONDITIONING",
"type": "CONDITIONING",
"links": [
2
],
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "CLIPTextEncode"
},
"widgets_values": [
"closeup photo of a woman wearing a white spring dress in a garden\n\nhigh quality, diffuse light, highly detailed, 4k"
],
"color": "#232",
"bgcolor": "#353"
},
{
"id": 5,
"type": "CLIPTextEncode",
"pos": [
710,
777
],
"size": {
"0": 400,
"1": 160
},
"flags": {},
"order": 8,
"mode": 0,
"inputs": [
{
"name": "clip",
"type": "CLIP",
"link": 6
}
],
"outputs": [
{
"name": "CONDITIONING",
"type": "CONDITIONING",
"links": [
3
],
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "CLIPTextEncode"
},
"widgets_values": [
"blurry, malformed, distorted, naked"
],
"color": "#322",
"bgcolor": "#533"
},
{
"id": 3,
"type": "EmptyLatentImage",
"pos": [
892,
994
],
"size": {
"0": 210,
"1": 110
},
"flags": {},
"order": 4,
"mode": 0,
"outputs": [
{
"name": "LATENT",
"type": "LATENT",
"links": [
4
],
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "EmptyLatentImage"
},
"widgets_values": [
512,
512,
1
]
},
{
"id": 10,
"type": "LoadImage",
"pos": [
180,
-50
],
"size": [
390.73314006347687,
482.01744979858427
],
"flags": {},
"order": 5,
"mode": 0,
"outputs": [
{
"name": "IMAGE",
"type": "IMAGE",
"links": [
77
],
"shape": 3,
"slot_index": 0
},
{
"name": "MASK",
"type": "MASK",
"links": null,
"shape": 3
}
],
"properties": {
"Node name for S&R": "LoadImage"
},
"widgets_values": [
"cwf_portrait.jpg",
"image"
],
"color": "#223",
"bgcolor": "#335"
},
{
"id": 14,
"type": "LoraLoaderModelOnly",
"pos": [
770,
410
],
"size": [
210,
82
],
"flags": {},
"order": 6,
"mode": 0,
"inputs": [
{
"name": "model",
"type": "MODEL",
"link": 15
}
],
"outputs": [
{
"name": "MODEL",
"type": "MODEL",
"links": [
16
],
"shape": 3,
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "LoraLoaderModelOnly"
},
"widgets_values": [
"ip-adapter-faceid_sd15_lora.safetensors",
0.6
],
"color": "#223",
"bgcolor": "#335"
},
{
"id": 1,
"type": "KSampler",
"pos": [
1380,
620
],
"size": {
"0": 240,
"1": 260
},
"flags": {},
"order": 10,
"mode": 0,
"inputs": [
{
"name": "model",
"type": "MODEL",
"link": 78
},
{
"name": "positive",
"type": "CONDITIONING",
"link": 2
},
{
"name": "negative",
"type": "CONDITIONING",
"link": 3
},
{
"name": "latent_image",
"type": "LATENT",
"link": 4
}
],
"outputs": [
{
"name": "LATENT",
"type": "LATENT",
"links": [
7
],
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "KSampler"
},
"widgets_values": [
2,
"fixed",
25,
6,
"dpmpp_2m",
"karras",
1
]
},
{
"id": 22,
"type": "PreviewImage",
"pos": [
1861,
542
],
"size": [
527.0097535195305,
548.0778934960933
],
"flags": {},
"order": 12,
"mode": 0,
"inputs": [
{
"name": "images",
"type": "IMAGE",
"link": 72
}
],
"properties": {
"Node name for S&R": "PreviewImage"
}
},
{
"id": 11,
"type": "IPAdapterApply",
"pos": [
1060,
240
],
"size": {
"0": 210,
"1": 258
},
"flags": {},
"order": 9,
"mode": 0,
"inputs": [
{
"name": "ipadapter",
"type": "IPADAPTER",
"link": 10
},
{
"name": "clip_vision",
"type": "CLIP_VISION",
"link": 11
},
{
"name": "image",
"type": "IMAGE",
"link": 77
},
{
"name": "model",
"type": "MODEL",
"link": 16
},
{
"name": "attn_mask",
"type": "MASK",
"link": null
}
],
"outputs": [
{
"name": "MODEL",
"type": "MODEL",
"links": [
78
],
"shape": 3,
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "IPAdapterApply"
},
"widgets_values": [
0.9,
0,
"original",
0,
1,
false
],
"color": "#223",
"bgcolor": "#335"
}
],
"links": [
[
2,
4,
0,
1,
1,
"CONDITIONING"
],
[
3,
5,
0,
1,
2,
"CONDITIONING"
],
[
4,
3,
0,
1,
3,
"LATENT"
],
[
5,
2,
1,
4,
0,
"CLIP"
],
[
6,
2,
1,
5,
0,
"CLIP"
],
[
7,
1,
0,
6,
0,
"LATENT"
],
[
8,
7,
0,
6,
1,
"VAE"
],
[
10,
9,
0,
11,
0,
"IPADAPTER"
],
[
11,
12,
0,
11,
1,
"CLIP_VISION"
],
[
15,
2,
0,
14,
0,
"MODEL"
],
[
16,
14,
0,
11,
3,
"MODEL"
],
[
72,
6,
0,
22,
0,
"IMAGE"
],
[
77,
10,
0,
11,
2,
"IMAGE"
],
[
78,
11,
0,
1,
0,
"MODEL"
]
],
"groups": [],
"config": {},
"extra": {},
"version": 0.4
}
examples/FaceID_enhanced.json
+875
View File
@@ -0,0 +1,875 @@
{
"last_node_id": 30,
"last_link_id": 77,
"nodes": [
{
"id": 6,
"type": "VAEDecode",
"pos": [
1670,
620
],
"size": {
"0": 140,
"1": 50
},
"flags": {},
"order": 15,
"mode": 0,
"inputs": [
{
"name": "samples",
"type": "LATENT",
"link": 7
},
{
"name": "vae",
"type": "VAE",
"link": 8
}
],
"outputs": [
{
"name": "IMAGE",
"type": "IMAGE",
"links": [
72
],
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "VAEDecode"
}
},
{
"id": 7,
"type": "VAELoader",
"pos": [
1380,
920
],
"size": {
"0": 240,
"1": 60
},
"flags": {},
"order": 0,
"mode": 0,
"outputs": [
{
"name": "VAE",
"type": "VAE",
"links": [
8
],
"shape": 3,
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "VAELoader"
},
"widgets_values": [
"vae-ft-mse-840000-ema-pruned.safetensors"
]
},
{
"id": 12,
"type": "InsightFaceLoader",
"pos": [
380,
240
],
"size": {
"0": 210,
"1": 58
},
"flags": {},
"order": 1,
"mode": 0,
"outputs": [
{
"name": "CLIP_VISION",
"type": "CLIP_VISION",
"links": [
11
],
"shape": 3,
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "InsightFaceLoader"
},
"widgets_values": [
"CPU"
],
"color": "#223",
"bgcolor": "#335"
},
{
"id": 9,
"type": "IPAdapterModelLoader",
"pos": [
300,
130
],
"size": {
"0": 300,
"1": 60
},
"flags": {},
"order": 2,
"mode": 0,
"outputs": [
{
"name": "IPADAPTER",
"type": "IPADAPTER",
"links": [
10
],
"shape": 3,
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "IPAdapterModelLoader"
},
"widgets_values": [
"ip-adapter-faceid_sd15.bin"
],
"color": "#223",
"bgcolor": "#335"
},
{
"id": 19,
"type": "PrepImageForClipVision",
"pos": [
700,
80
],
"size": [
243.60000610351562,
106
],
"flags": {},
"order": 11,
"mode": 0,
"inputs": [
{
"name": "image",
"type": "IMAGE",
"link": 29
}
],
"outputs": [
{
"name": "IMAGE",
"type": "IMAGE",
"links": [
30
],
"shape": 3,
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "PrepImageForClipVision"
},
"widgets_values": [
"LANCZOS",
"center",
0
],
"color": "#223",
"bgcolor": "#335"
},
{
"id": 17,
"type": "CLIPVisionLoader",
"pos": [
730,
-40
],
"size": [
210,
58
],
"flags": {},
"order": 3,
"mode": 0,
"outputs": [
{
"name": "CLIP_VISION",
"type": "CLIP_VISION",
"links": [
70
],
"shape": 3,
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "CLIPVisionLoader"
},
"widgets_values": [
"IPAdapter_image_encoder_sd15.safetensors"
],
"color": "#223",
"bgcolor": "#335"
},
{
"id": 18,
"type": "IPAdapterModelLoader",
"pos": [
640,
-160
],
"size": {
"0": 300,
"1": 60
},
"flags": {},
"order": 4,
"mode": 0,
"outputs": [
{
"name": "IPADAPTER",
"type": "IPADAPTER",
"links": [
28
],
"shape": 3,
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "IPAdapterModelLoader"
},
"widgets_values": [
"ip-adapter-plus-face_sd15.safetensors"
],
"color": "#223",
"bgcolor": "#335"
},
{
"id": 22,
"type": "PreviewImage",
"pos": [
1880,
620
],
"size": {
"0": 380,
"1": 410
},
"flags": {},
"order": 16,
"mode": 0,
"inputs": [
{
"name": "images",
"type": "IMAGE",
"link": 72
}
],
"properties": {
"Node name for S&R": "PreviewImage"
}
},
{
"id": 2,
"type": "CheckpointLoaderSimple",
"pos": [
50,
700
],
"size": {
"0": 290,
"1": 100
},
"flags": {},
"order": 5,
"mode": 0,
"outputs": [
{
"name": "MODEL",
"type": "MODEL",
"links": [
15
],
"slot_index": 0
},
{
"name": "CLIP",
"type": "CLIP",
"links": [
5,
6
],
"slot_index": 1
},
{
"name": "VAE",
"type": "VAE",
"links": [],
"slot_index": 2
}
],
"properties": {
"Node name for S&R": "CheckpointLoaderSimple"
},
"widgets_values": [
"sd15/realisticVisionV51_v51VAE.safetensors"
]
},
{
"id": 1,
"type": "KSampler",
"pos": [
1380,
620
],
"size": {
"0": 240,
"1": 260
},
"flags": {},
"order": 14,
"mode": 0,
"inputs": [
{
"name": "model",
"type": "MODEL",
"link": 36
},
{
"name": "positive",
"type": "CONDITIONING",
"link": 2
},
{
"name": "negative",
"type": "CONDITIONING",
"link": 3
},
{
"name": "latent_image",
"type": "LATENT",
"link": 4
}
],
"outputs": [
{
"name": "LATENT",
"type": "LATENT",
"links": [
7
],
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "KSampler"
},
"widgets_values": [
0,
"fixed",
25,
6,
"dpmpp_2m",
"karras",
1
]
},
{
"id": 11,
"type": "IPAdapterApply",
"pos": [
730,
240
],
"size": {
"0": 210,
"1": 258
},
"flags": {},
"order": 12,
"mode": 0,
"inputs": [
{
"name": "ipadapter",
"type": "IPADAPTER",
"link": 10
},
{
"name": "clip_vision",
"type": "CLIP_VISION",
"link": 11
},
{
"name": "image",
"type": "IMAGE",
"link": 77
},
{
"name": "model",
"type": "MODEL",
"link": 16
},
{
"name": "attn_mask",
"type": "MASK",
"link": null
}
],
"outputs": [
{
"name": "MODEL",
"type": "MODEL",
"links": [
35
],
"shape": 3,
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "IPAdapterApply"
},
"widgets_values": [
0.9,
0,
"original",
0,
1,
false
],
"color": "#223",
"bgcolor": "#335"
},
{
"id": 4,
"type": "CLIPTextEncode",
"pos": [
710,
570
],
"size": {
"0": 400,
"1": 160
},
"flags": {},
"order": 9,
"mode": 0,
"inputs": [
{
"name": "clip",
"type": "CLIP",
"link": 5
}
],
"outputs": [
{
"name": "CONDITIONING",
"type": "CONDITIONING",
"links": [
2
],
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "CLIPTextEncode"
},
"widgets_values": [
"closeup photo of a woman wearing a white spring dress in a garden\n\nhigh quality, diffuse light, highly detailed, 4k"
],
"color": "#232",
"bgcolor": "#353"
},
{
"id": 5,
"type": "CLIPTextEncode",
"pos": [
710,
777
],
"size": {
"0": 400,
"1": 160
},
"flags": {},
"order": 10,
"mode": 0,
"inputs": [
{
"name": "clip",
"type": "CLIP",
"link": 6
}
],
"outputs": [
{
"name": "CONDITIONING",
"type": "CONDITIONING",
"links": [
3
],
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "CLIPTextEncode"
},
"widgets_values": [
"blurry, malformed, distorted, naked"
],
"color": "#322",
"bgcolor": "#533"
},
{
"id": 3,
"type": "EmptyLatentImage",
"pos": [
892,
994
],
"size": {
"0": 210,
"1": 110
},
"flags": {},
"order": 6,
"mode": 0,
"outputs": [
{
"name": "LATENT",
"type": "LATENT",
"links": [
4
],
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "EmptyLatentImage"
},
"widgets_values": [
512,
512,
1
]
},
{
"id": 16,
"type": "IPAdapterApply",
"pos": [
1070,
180
],
"size": {
"0": 210,
"1": 258
},
"flags": {},
"order": 13,
"mode": 0,
"inputs": [
{
"name": "ipadapter",
"type": "IPADAPTER",
"link": 28
},
{
"name": "clip_vision",
"type": "CLIP_VISION",
"link": 70,
"slot_index": 1
},
{
"name": "image",
"type": "IMAGE",
"link": 30
},
{
"name": "model",
"type": "MODEL",
"link": 35
},
{
"name": "attn_mask",
"type": "MASK",
"link": null
}
],
"outputs": [
{
"name": "MODEL",
"type": "MODEL",
"links": [
36
],
"shape": 3,
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "IPAdapterApply"
},
"widgets_values": [
0.35000000000000003,
0.33,
"original",
0,
1,
false
],
"color": "#223",
"bgcolor": "#335"
},
{
"id": 10,
"type": "LoadImage",
"pos": [
-150,
-50
],
"size": [
390.73314006347687,
482.01744979858427
],
"flags": {},
"order": 7,
"mode": 0,
"outputs": [
{
"name": "IMAGE",
"type": "IMAGE",
"links": [
29,
77
],
"shape": 3,
"slot_index": 0
},
{
"name": "MASK",
"type": "MASK",
"links": null,
"shape": 3
}
],
"properties": {
"Node name for S&R": "LoadImage"
},
"widgets_values": [
"cwf_portrait.jpg",
"image"
],
"color": "#223",
"bgcolor": "#335"
},
{
"id": 14,
"type": "LoraLoaderModelOnly",
"pos": [
440,
410
],
"size": [
210,
82
],
"flags": {},
"order": 8,
"mode": 0,
"inputs": [
{
"name": "model",
"type": "MODEL",
"link": 15
}
],
"outputs": [
{
"name": "MODEL",
"type": "MODEL",
"links": [
16
],
"shape": 3,
"slot_index": 0
}
],
"properties": {
"Node name for S&R": "LoraLoaderModelOnly"
},
"widgets_values": [
"ip-adapter-faceid_sd15_lora.safetensors",
0.6
],
"color": "#223",
"bgcolor": "#335"
}
],
"links": [
[
2,
4,
0,
1,
1,
"CONDITIONING"
],
[
3,
5,
0,
1,
2,
"CONDITIONING"
],
[
4,
3,
0,
1,
3,
"LATENT"
],
[
5,
2,
1,
4,
0,
"CLIP"
],
[
6,
2,
1,
5,
0,
"CLIP"
],
[
7,
1,
0,
6,
0,
"LATENT"
],
[
8,
7,
0,
6,
1,
"VAE"
],
[
10,
9,
0,
11,
0,
"IPADAPTER"
],
[
11,
12,
0,
11,
1,
"CLIP_VISION"
],
[
15,
2,
0,
14,
0,
"MODEL"
],
[
16,
14,
0,
11,
3,
"MODEL"
],
[
28,
18,
0,
16,
0,
"IPADAPTER"
],
[
29,
10,
0,
19,
0,
"IMAGE"
],
[
30,
19,
0,
16,
2,
"IMAGE"
],
[
35,
11,
0,
16,
3,
"MODEL"
],
[
36,
16,
0,
1,
0,
"MODEL"
],
[
70,
17,
0,
16,
1,
"CLIP_VISION"
],
[
72,
6,
0,
22,
0,
"IMAGE"
],
[
77,
10,
0,
11,
2,
"IMAGE"
]
],
"groups": [],
"config": {},
"extra": {},
"version": 0.4
}
examples/face_id_wf.jpg
BIN
View File
Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 186 KiB