After Detailer Extension In 9 Minutes - Stable Diffusion Tutorial (Automatic1111)

After Detailer can take low quality faces and make them high quality automatically, without manual inpainting.
-LINKS-
(When using an affiliate link, I earn a commission which is a fantastic way to support the channel)
➤ GENERAL ACCESSORIES - amzn.to/3Jl9zpe
➤ NVIDIA GRAPHICS CARDS - amzn.to/3TZNGRn
➤ AMD GRAPHICS CARDS - amzn.to/3PYIaNN
➤ INTEL CPU - amzn.to/3vRlYya
➤ RAM - amzn.to/3Jm53Hf
➤ CORSAIR VENGEANCE 16GB - amzn.to/3PY8JTm
➤ CORSAIR VENGEANCE 32GB - amzn.to/4aRTwLx
➤ CORSAIR VENGEANCE 64GB - amzn.to/3Q1NboQ
➤ GENERAL MONITORS - amzn.to/3xEk6td
➤GENERAL KEYBOARDS AND MOUSES - amzn.to/4aAaEpp
➤GAMING KEYBOARDS AND MOUSE - amzn.to/49wpYly
Support the Channel:
➤ Patreon: / bitesizedgenius
➤ Buy Me Coffee: bmc.link/bitesizedgenius
Links:
➤ Where Files Go: • Explaining Stable Diff...
➤ Installing Extensions: • Explaining Stable Diff...
➤ XYZ Plot: • Explaining XYZ Plot In...
➤ After Detailer: github.com/Bing-su/adetailer
Chapters
0:00 - Introduction
0:30 - Enable After Detailer
0:40 - Two Tabs & 2 Models
0:54 - Two Prompt Boxes
1:29 - Yolo Models
1:54 - Media Pipe Models
2:21 - Hand Models
2:44 - Full Body Model
2:57 - Detection Confidence Threshold
3:31 - Mask Only The Top K Largest
3:56 - Mask Min and Max Ratio
4:14 - Mask X and Y Offset
4:30 - Mask Erosion / Dilation
4:47 - Mask Merge Modes
5:12 - Inpaint Mask Blur
5:32 - Inpaint Only Mask
5:53 - Inpaint Only Masked Padding Pixels
6:10 - Use Seperate
6:44 - There Are Also Standard Settings
7:00 - Inpaint Denoising Strength
7:14 - Inpaint Width and Height
7:31 - Noise Muliplier
7:51 - Restore Faces After Adetailer
8:05 - Control Net
8:32 - Conclusion

[@magnusberge]

00:42 you can have more tabs than 1st and 2nd, in the settings tab at the top of AUTO 11 11 there is now a Adetailer setting that lets you add more tabs like 3rd and 4th etc.

[@BitesizedGenius]

Gem of a comment, was just wondering why something like this didn't exist!

[@zeror6590]

​@BitesizedGenius bro I need ur help, which setting do I need to change so that I can get a new face and not the same always? When i change the prompt or not it somehow gives the same face always once it generates it for once. So how to get new faces while still being effective?

[@BitesizedGenius]

@@zeror6590 Denoising strength in adetailer, try setting it to 0.7 or 1 when using the face model.

[@zeror6590]

​@BitesizedGenius alright thanks! Subscribing also to support you :)

[@Slav4o911]

@@zeror6590 You can type different text for the face and use embeddings and Lora there. By default Adetailer uses your original prompt for the face, but you can type whatever prompt you want. I've found it works very well with text embeddings. Adetailer is the best way to make a better face without having to do face swaps and other more complex things.

[@Xungames]

So much information in less than 10 minutes. I'll have to watch it a few times to understand everything. Thanks for the chapters, it's a huge time saver. 👍

[@Starlingstudio]

Thank you, Very great video

[@Numb_]

I feel like your conclusion should include your recommended settings based on common use cases

[@TheKamrenB]

Thank you. I have read that face_yolo models are anime-focused and mediapipe_face models are for photos/realistic faces.

[@BitesizedGenius]

Ah that would make sense!

[@polystormstudio]

It's nice to finally see an in-depth explanation of this extension. However in looking at the comparisons, they're so miniscule I don't see why I would go any further than the default settings.

[@BitesizedGenius]

Most times the default settings are fine. I only really change the inpainting strength and sometimes the detection settings if I have background characters.

[@How-xw1ws]

Anyone know how to increase the automatic in paint size when using the mediapipe_face_mesh_eyes_only? I find it is able to find where the eyes are but it doesn't paint in enough of the eyes. It only covers like 30% of the eyes since my character has really large eyes. I want it to basically use a larger brush when covering the area where it thinks the eyes are

[@darkskinnedpimp]

The lady in the far left of the thumbnail is going to give me nightmares.. primarily just commenting for the algorithm cuz I like your videos 👍 love the video! #Amazing

[@grzesiektg]

does it work with SDXL?

[@BitesizedGenius]

Yep, it's an extension so should work with SDXL in Automatic1111

[@ctrlartdel]

For some reason, nothing is making sense to me. Maybe it’s the photos you were using. Nothing is different about them. I don’t know. I just heard mask bunch of times.

[@vitalis]

If we could get a way to really fix hands….

[@sethcarter4498]

He mentioned embeddings (textual inversion) and negative prompts, which put in a ton of work to do just this. "BadHands" is an example of a popular textural inversion to fix hands.

[@DNozz777]

I hate to say it because I know video editing is a lot of work, but it's kinda a terrible video. It was too vague and fast. 9 minutes just isn't enough time for a viewer to grasp these concepts. It would also help if you didn't sound like you were reading off a paper. I simply couldn't follow. Sorry for the criticism, I hope you use to make improvements. Anyway, thanks for putting it out there!

[@ctrlartdel]

I think I’m more confused

[@ctrlartdel]

You rock though dude thanks for all the help! But for some reason, the words you’re stringing together, aren’t making sense to me