Workflow description :
The aim of this workflow is to generate video using the face from an existing photo in a simple window.
Resources you need:
馃搨Files :
WAN2.1
Recommendation :
24 gb Vram: Q8_0
16 gb Vram: Q5_K_S
<12 gb Vram: Q4_K_S
I2V Quant Model: Wan2.1-I2V-14B-480P-gguf or Wan2.1-I2V-14B-720P-gguf
In models/diffusion_models
CLIP: umt5_xxl_fp8_e4m3fn_scaled.safetensors
in models/clip
CLIP-VISION: clip_vision_h.safetensors
in models/clip_vision
VAE: wan_2.1_vae.safetensors
in models/vae
FLUX
GGUF_Model: FLUX.1-dev-gguf
"flux1-dev-Q8_0.gguf" in ComfyUI\models\unet
GGUF_clip: t5-v1_1-xxl-encoder-gguf
"t5-v1_1-xxl-encoder-Q8_0.gguf" in \ComfyUI\models\clip
Text encoder: ViT-L-14-TEXT-detail-improved-hiT-GmP-TE-only-HF.safetensors
"ViT-L-14-GmP-ft-TE-only-HF-format.safetensors" in \ComfyUI\models\clip
VAE: ae.safetensors
"ae" in \ComfyUI\models\vae
FLUX PuLID : pulid_flux_v0.9.0.safetensors
"pulid_flux_v0.9.0" in \ComfyUI\models\pulid
ANY upscale model (depreciated):
Realistic : RealESRGAN_x4plus.pth
Anime : RealESRGAN_x4plus_anime_6B.pth
in models/upscale_models
馃摝Custom Nodes :
PuLID need your python to have Insightface :
Check your python version :
for windows portable version : (the path depends on where you unzipped ComfyUI)
Download Insightface whl which corresponds here : Assets/Insightface
(Here my local python is on 310 and mobile version in 312)
Then install all prerequisites and insightface :
python.exe -m pip install --use-pep517 facexlibpython.exe -m pip install git+https://github.com/rodjjo/filterpy.git
python.exe -m pip install onnxruntime==1.19.2 onnxruntime-gpu==1.15.1 insightface-0.7.3-cp310-cp310-win_amd64.whl
Description
bugfix with PuLID link
FAQ
Comments (22)
First off, thanks for your work. Impressive stuff.
I'm not even sure why I got this workflow, I have all the files but what exactly is the purpose of this workflow?
This workflow allows you to reproduce a face in a video
Is this a flux pluid wf that feeds into a i2v wan wf?
Yes
ok, I have this working now.. However, I enter a Caucasian woman and the workflow generates an asian woman. I thought the idea of this was to keep the face the same?
With the 1.1 version? As you can see in my exemple i use a caucasian girl and have a caucasian woman in the video
The preview image is an asian girl too?
@UmeAiRT聽Ok, yes I had the old version. I updated to 1.1, now its working nice. Thanks for the effort on this. Yes, the resemblance is pretty good. I definitely will be using the workflow. The resemblance isn't perfect, but it is pretty good overall.
@trashkollector175聽It is possible to play a little with the settings in the PuLID node for a better resemblance
not working fine...
Great workflow! Any plans to add WAN lora?
You can easily add lora by taking example from my two other workflows for WAN
I have been using this for a few days now.. this workflow is great.. kudos to the author.
Thanks !
Thank you for this, and thank you even more for your scripts. I was struggling to install the missing nodes in comfy but your script resolved it for me and the guide was easy to follow, I'll upload some videos after playing around with it for a bit.
Thanks for your feedback.
@UmeAiRT聽I like how you have these workflows set up, but for some reason after I resize the image I can never get it set back to 720, I always have to use 512x768. Is it possible to keep what you have but also add the option to type the exact size if needed?
Sometimes I get blurry teeth as well, I'm going to try and change the upscaler model to see if that helps fix the issue, but it might end up needing some kind of facedetailer node.
If for some reason a facedetailer node is needed is it worth also adding one for the hands as well?
@synalon973聽To choose the size in exact numbers, simply double-click on the number and an editor will open.
@UmeAiRT聽When I double click it doesn't give me an editor, but its a minor issue anyway.
RuntimeError: Can't access transformer_options, this requires ComfyUI nightly version from Mar 14, 2025 or later