ComfyUI Refacer - 18th May Edition
Added 2024-05-18 15:20:01 +0000 UTCWhile various things keep coming out which use the research-only "insightface" models, I figured why not combine two insightface-free options together? On top of that, maybe see how quickly I can make that cartoon or painted face into a more realistic one? We've had a load of new nodes come out recently, so now is as good a time as any to slap a load into a single workflow.
Below you'll find the ComfyUI workflows for both real and anime preset "Refacer" examples.
You'll need a whole load of custom nodes installed, along with the various models, which you can download from the links below, if you don't already have them:
Hyper-SDXL = https://huggingface.co/ByteDance/Hyper-SD
IP Adapter = https://huggingface.co/h94/IP-Adapter
Photomaker = https://huggingface.co/TencentARC/PhotoMaker
The easiest way to install any missing nodes is through ComfyUI manager, with the custom node github pages providing additional information, if required.
Enjoy refacing! :)
Video guide - https://youtu.be/r7Iz8Ps7R2s
Update - 19th May
Added a few more presets:
Painting
Sketch
Sketch with pose
Update - 21st May
Still having too much fun with this ;) Some slightly different presets for "Sketch" this time (v1). I was seeing how high I could get the weight value, so it's higher than in v0. Usually that adds too much colour. However, by keeping really tight "start" and "end" values, much of the face fidelity benefits from the IP Adapter face model can kept without the colour creeping in. "end at" is about as high as I think it can go, but I've only tested a limited set of images. If you're getting colour, lowering "end at" a tiny amount should do the trick. Happy AI sketching!
Comments
Ahh, makes sense. I've noticed a few people doing that and wasn't sure the purpose when they are newer workflows but I can see that. Thanks!
GeoMage
2024-06-06 12:12:06 +0000 UTCI personally prefer simplicity over obfuscation :)
Nerdy Rodent
2024-06-06 10:09:16 +0000 UTCJust curious...is there a reason you're using separate IPAdapter and Clip Vision loaders rather than the new unified loader?
GeoMage
2024-06-02 19:18:44 +0000 UTCThank you!
Rahil bin ali
2024-06-02 04:26:42 +0000 UTCThe clip vision model is this one - https://huggingface.co/h94/IP-Adapter/tree/main/models/image_encoder
Nerdy Rodent
2024-06-01 18:53:42 +0000 UTCUPDATE: Managed to install everything just getting this error now when i queue prompt - "Missing CLIPVision model" . I have copied your exact wrokflow through the video all things are in their respective folders. Plz help me out
Rahil bin ali
2024-06-01 14:56:10 +0000 UTCcould you please tell me which versions to download its a bit confusing .
Rahil bin ali
2024-06-01 09:33:49 +0000 UTCHyper-SDXL = https://huggingface.co/ByteDance/Hyper-SD IP Adapter = https://huggingface.co/h94/IP-Adapter Photomaker = https://huggingface.co/TencentARC/PhotoMaker
Rahil bin ali
2024-06-01 09:33:00 +0000 UTCOh more workflows. Thanks very fun
Clayfacer
2024-05-22 06:15:31 +0000 UTClovely, like the compactness & versatility of your workflows, well done :)
Christian G.
2024-05-21 19:30:04 +0000 UTC
