ComfyUI Impact Pack - Q&A: Detailer Options (guide_size, guide_size_for, crop_factor, force_inpaint)

preview_player
Показать описание
This video explains guide_size, guide_size_for, crop_factor, and force_inpaint used in detailer.
Рекомендации по теме
Комментарии
Автор

Thank you for making these videos, they help a lot with trying to understand how these nodes work. Would you be willing to add the workflow json file ?

BYTEcann
Автор

If only others realised the power they can wield by getting COMFY!! LOL. Actually I get why this could be so daunting to others when they look at all the boxes and tangled spaghetti. somw how this speaks to me. I have left A1111 well and truly in the dust to get COMFY!!! Thank you so much for the workflows!!!! for a noob like me who loves to learn. O..M..G!!! SUBBED!!!!😍

yiluwididreaming
Автор

Thank you. The detailer just skipped when fed with higher resolution image. Now I know what to tinker to make it work.

Najdmie
Автор

Hi, First of all you are amazing, and I really think you are a genius :D
Just out of curiosity - you said in 2:23 " you can observe..." can you please provide some insight on how come the upscale was 2.47?, I mean why wasn't it was 2.82X304 - assuming the guide_size_for is still set to bbox...I hope I'm clear...

lhovav
Автор

Great Tips, thanks for the videos, how do you change the path where images are stored to another directory?

RamonGuthrie
Автор

thanks for the videos. May I ask what works are available for the parameter drop_size? I have been looking for it for a long time, but I haven't found the relevant instructions. Thank you.

GUOREE
Автор


This workflow is based on faceswap.json on above repository.

Please note that the videos were created using a version that is currently scheduled for an update, and workflows will also be re-uploaded along with the updates. The workflows in the repository may currently have some version compatibility issues.

drltdata
Автор

I use the face detailer (DetailerDebug (SEGS) with samdetector (combined) and BBOX Detector (SEGS) to be specific) to improve faces. Sometimes the mask is a little bit too large or the transition is too obvious. I play with the dilation parameter until I get the desired result but I was wondering. Is it possible to edit the mask while using the SEGS workflow?

kofteburger
Автор

The MMdetDetector provider has only one model . an anime one. Is there a way possible to add new models to this?

musicandhappinessbyjo
Автор

I like impact pact very much. For face detailer, is there other alternative than mmdet_anime-face_yolov3.pth than I can download and use? That model give very anime-ish face which I may not want for men. Though Face Bbox(bounding box) + Person silhouette segmentation workflow work great as a workaround.

Najdmie