[Workflow] Kolors_V Try-On [Advance] [Ver.20241101]
🕑 Added 2024-11-01 14:00:18 +0000 UTC
Tutorial Video: https://youtu.be/tdbntcnE79A
This workflow are going to continue with update, the next version, I will be try to enhance the Try-On success rate. I think that is all we need to focus for V Try On workflow.
Below are list of models and custom node that need for running Kolors Diffusion:
Here's the model your need :
https://huggingface.co/spaces/Kwai-Kolors/Kolors-Virtual-Try-On
https://github.com/Kwai-Kolors/Kolors/blob/master/imgs/Kolors_paper.pdf
https://huggingface.co/Kwai-Kolors/Kolors
https://huggingface.co/Kwai-Kolors/Kolors-IP-Adapter-Plus
For Custom Node:
I built the workflow base on this custom node
https://github.com/MinusZoneAI/ComfyUI-Kolors-MZ
Implementation of Kolors on ComfyUI
UNET model download
The unet model is placed in the models/unet/ folder
Model homepage: https://huggingface.co/Kwai-Kolors/Kolors
Download link: https://huggingface.co/Kwai-Kolors/Kolors/resolve/main/unet/diffusion_pytorch_model.fp16.safetensors
ChatGLM3 model download
The chatglm3 is placed in the models/LLM/ folder
Model homepage: https://huggingface.co/Kijai/ChatGLM3-safetensors
Download link: https://huggingface.co/Kijai/ChatGLM3-safetensors/resolve/main/chatglm3-fp16.safetensors
IPAdapter implementation is recommended to use cubiq/ComfyUI_IPAdapter_plus
The faceid related nodes have been supported in it, and the IPAdapter implementation needs to update ComfyUI_IPAdapter_plus to the latest version
Official IP-Adapter-Plus model download link
Model homepage: https://huggingface.co/Kwai-Kolors/Kolors-IP-Adapter-Plus
https://huggingface.co/Kwai-Kolors/Kolors-IP-Adapter-Plus/resolve/main/ip_adapter_plus_general.bin Download to models/ipadapter/
https://huggingface.co/Kwai-Kolors/Kolors-IP-Adapter-Plus/resolve/main/image_encoder/pytorch_model.bin Download to models/clip_vision/
Official Kolors-IP-Adapter-FaceID-Plus model download link
Model homepage: https://huggingface.co/Kwai-Kolors/Kolors-IP-Adapter-FaceID-Plus
https://huggingface.co/Kwai-Kolors/Kolors-IP-Adapter-FaceID-Plus/resolve/main/ipa-faceid-plus.bin Download to models/ipadapter/
https://huggingface.co/Kwai-Kolors/Kolors-IP-Adapter-FaceID-Plus/resolve/main/clip-vit-large-patch14-336/pytorch_model.bin Download to models/clip_vision/
https://huggingface.co/MonsterMMORPG/tools/resolve/main/antelopev2.zip Download and unzip to models/insightface/models/antelopev2/*.onnx
Official ControlNet model download link
Model homepage(Depth): https://huggingface.co/Kwai-Kolors/Kolors-ControlNet-Depth
Model homepage(Canny): https://huggingface.co/Kwai-Kolors/Kolors-ControlNet-Canny
Model homepage(Pose): https://huggingface.co/Kwai-Kolors/Kolors-ControlNet-Pose
Kolors-Inpainting model download link
Model homepage: https://huggingface.co/Kwai-Kolors/Kolors-Inpainting
https://huggingface.co/Kwai-Kolors/Kolors-Inpainting/resolve/main/unet/diffusion_pytorch_model.safetensors Download to models/unet/
Comments
Benjamin Law
Actually, I used the Flux clip fp8 before, and it work with this.
Error_404_unknown
Hi Benji, I'm so sorry to ask but I'm totally lost with the Kolors Ip adapter kolors\model(dot)bin. I downloaded all kolors models in comfy but this model doesn't show in the load clip vison node. I went to the huggingface and downloaded all models as well and placed them in the appropriate folders but still I can't get this model to show I only see clip-vit-h, clip-vit-large and clip-vit-big. Do you have any suggestions? Sorry if this is vague I can email with more details.
Benjamin Law
Comfyui i use poetable version, which it have its own Python library to run. Other AI tools I use Conda for each of it. Most often use , like OpenWebUI connect with Ollama, it use Docker image. So it is totally isolated.
modiin wrestler
Thank you for the amazing content you share—it’s been incredibly helpful! I noticed you mentioned working with multiple projects and workflows, and I was curious about how you manage your Python environments. Do you use separate Conda environments for each workflow, or do you keep everything in a single environment? For example, I often encounter issues where one workflow (like ComfyUI-LivePortraitKJ) fails due to dependencies (e.g., insightface) whenever I install a new workflow. I'm wondering if using separate environments might help avoid these conflicts, or if there’s a recommended approach for handling multiple workflows in one environment. Thanks for any advice you can share!
Benjamin Law
thanks, let me know if you have any issue or feedback for improvement. I am keep updating some workflows, that have potential. If you see there's a version number on the title, those will be getting update if possible.
aprisma
Amzing workflow. And very good - and this is the most important thing - description what nodes/model used and where to find them. Thank you so much.
Benjamin Law
VAE are using SDXL VAE