ComfyUI: Face Detailer (Workflow Tutorial)

This tutorial includes 4 Comfy UI workflows using Face Detailer. The workflow tutorial focuses on Face Restore using Base SDXL & Refiner, Face Enhancement (Graphic to Photorealistic & Vice Versa), and Facial Details (including hair styling). Unique techniques are used to automate the workflow for auto-detection, selection, and masking.

------------------------

------------------------

TimeStamps:

0:00 Intro.
0:46 Requirements.
2:08 Face Restore.
10:46 Face Restore Extra.
12:26 Face Enhancement.
15:14 Face Details.
26:08 Upscaling.

[ericgoodman]

This tutorial is probably the best I've ever seen on any subject. I will still have to watch it many times to absorb the information.

[OnePikki]

Thank you very much to give a information about the detailer settings do work! Great vid!

[johnriperti]

Beautiful, so much info, thanks and your voice is soothing

[drltdata]

This tutorial video is the most detailed among the ones I've watched.

[ProJHTV]

Perfect tutorial. Can't wait for your next video.

[nodewizard]

Great guide. Very detailed and step-by-step. Thanks.

[Darkwing]

Excellent video! Really liked the detailed explanations of what the different settings do.

[HappyRadio_]

Thank you! with this i was able to add this to my animatediff batch list omg, thanks.

[Mootai]

wow this video is so full of information. Thanks a lot !

[rolgnav]

I don't usually comment, but wow, this tutorial explained a lot. Thank you!

[Nrek_AI]

you've given us so much info here. Thank you so much! I learned so much

[MannyGonzalez]

OMG, this was an insane learning experience. Thank you for sharing!!

[banzaipiegaming]

Absolutely solid videos all around on this channel. Something I am having the upmost difficult finding anywhere online are taking art and converting them to photorealistic images. There are too many tutorials that show how to do photo to art style using control nets and IP adapters, but nothing for the inverted form of it. I've also asked around from more experienced people and they don't know how to change art styles to photos either. If it is possible and you can figure that out, a video on that would be awesome.

[p_p]

i felt stupid watching this video. thnaks for sharing your knowldge

[bxwppbh]

Hey, I like the workflow. However, the program is reporting an error: TypeError: got an unexpected keyword argument 'resolution'

[artisans]

The use of the word "too" many times is so important. It gives not only the how but also the why. The how doesn't teach, the why teaches.

[julienblanchon]

For zero shot segmentation, did you see massive difference between SAM+Grounding DINO vs ClipSeg ?

[xd]

Hey, I like the workflow. However, could you share a link to the repo where we can dl the bbox/segm models? I can't find all of those on the original github repo. Thanks :)

[runebinder]

Hi, great tutorial and got it to work with the ClipTextEncodeSDXL for the positive, when I replaced that with the Blip Analyse Image you showed afterwards, I get this error: "Error occurred when executing BLIP Analyze Image: The size of tensor a (6) must match the size of tensor b (36) at non-singleton dimension 0". Very new to ComfyUI and Stable Diffusion in general, any ideas where I'm going wrong? From what I can see online it's expecting 2 values to be the same but they're not and that's throwing up the error. No idea what those values are though.

[tamtamX-cqor]

Thank you for this awesome tutorial! With the mediapipe face mesh I’m running into issues where it errors very often when the face is to small in the image. But not taking about tiny. Still relatively large but still erroring.