Improve Faces, Hands, and Poses with ADetailer! 💥

🎉 The After Detailer (ADetailer) extension is built in Stable Diffusion to improve faces and hands in Automatic1111!

🎬 In this tutorial video, we will see how to download the ADetailer extension, and use it! We will learn how to master the main settings to generate higher-quality images.

💬 Social Media:
____________________________________________________________________

🤙🏻 Learn More:

📰Article:

📌 Links:

00:32 What is ADetailer
01:00 Install the ADetailer extension
01:37 Main GitHub page
01:56 Models you can use
03:37 My prompt
04:24 Enable the ADetailer (default settings)
05:50 Confidence Score
06:07 Detection Model Confidence Threshold
07:51 Inpainting Denoising Strength
08:49 Mask Min/Max Area Ratio
09:20 Positive and Negative Prompt
10:53 ControlNet
11:52 Other settings
12:28 Use two ADetailer models together
14:00 Final results

#aiart #stablediffusion #generativeart #stabilityai #stablediffusiontutorial
#sdxl #adetailer

[theartofrichardschouten]

It's great that you show different settings, and especially extreme settings. It paints a clear picture of the options. Thanks for your efforts!

[JemiloII]

I enjoy the amount of detail you give about this. I'm not using A1111, but coding stuff up. Reading the source code, not everything was quite clear how it was working. This explains it clearly. Much appreciated.

[thanksfernuthin]

I love Adetailer and use it for faces every time. It basically gives the image generator a lot more pixels to do the face. I used to have to inpaint to do that and now it does it automatically! Another good setting is "Mask only the top k largest (0 to disable)" in detection. It defaults to 0. If you set it to one it will only do the face that's the most prominent. Normally your main character. That way you don't need to dial in detection perfectly to keep it from doing a crowd in the background which would be awful. Not only does it take forever it gives everyone in the crowd the face of the character you described.
I tried hands for a while and realized it was doing harm half the time and improving once in a while. So I dropped it.
You don't have to play with SD for long before you realize asking for blue eyes or red lips gets you a bunch of stuff colored blue or red in the image. So it's great for face details.

[mohamedzewail]

Could you please make an inpainting and outpainting tutorial?, your way of simplifying SD topics is really good, so if you got the time please consider making it.

[aaronsj]

I've been using ADetailer for a few days and it's pretty cool. From my experience it seems to struggle with improving hands when the hands are holding something, but it does good when improving faces. I especially like being able to prompt specific details for the face separately and not have those details influence the rest of the image.

[Mavczers]

Thanks for explaining everything. I bought computer to have fun with Stable Diffusion. Now I feel like a child learning how to play piano by pressing each key and listening how it sounds.
Your guidance is very helpful - I feel more motivated to generate pic after a pic.

[ThisOrThat]

Thank you, thank you ssoo much for going over the sliders. Most videos I see with cover installation and highlights. Though rarely, the the fine tuning process.

[carolh.]

Thanks for the video! So far I've been using the face fix tool in SD to fix faces but I think I get better results with ADetailer. I gotta keep experimenting with this. Once again, thanks!

[johndebattista-qe]

Thank you to explain to me how this works I had a couple of problems with the extension roop realize I had to take some extension off from stable diffusion See if you can explain to them how this works Keep up with the good job you're doing Best regards
Laura Carnevali

[Vestu]

Great video! Just what I needed actually for a wallpaper project. Thanks!

[Yoshi]

Thank you for this detailed guide! :)
You did put alot of effort into this, very much appreciated! 🏆💯

[exstasis]

Good video. I'd be interested in a video tutorial for using Adetailer in ComfyUI, mainly since ComfyUI keeps the last intermittent images for each stage in a cache so that it doesnt need to redo anything if yhe original prompt never changed. It cuts down waiting time by a lot, especially for people like me who don't know what they are doing and are just experimenting blindly with settings.

[glennmclintock]

For faces, yep! All day long. It's pretty much crap for everything else. It'll occassionally fix hands, but it's been my experience that if it CAN'T fix the hands it makes the hand unfixable by any other method because it'll take a somewhat bad hand and turn it into an absolute twisted horror.

[Shabazza]

If a character has 2 visble hands, would you prompt in AD for 1 hand, or for both hands? And if the latter .... how?
Ah, nevermind. You can enable the "left to right" or "center to edge" mode and then use [SEP] and [SKIP] to prompt for each detected object.

[artymusoke]

Thanks for the tutorial. ive been using adetailer with no problem for afew months now. i recently made my first lora for braided hair, however the adetailer ads like an opaque square around the characters face in the final result. anyone know what could be causing this?

[oredzera]

It's possible to a mke a tutorial using this with Comfyui?

[OwOUwUOnO]

I was curious how ADetailer would work on a photo of my fat cat & it was hilarious 😂😂😂 for some reason ADetailer gave my cat human eyes & hands 😂😂😂😂 but mainly user error too

[abdurrahmanozkann]

ı ve a question, I wrote my prompt and stable generated my pic and then I activated Adetailer my first picture totally changed but your pic didn't change what is my wrong

[Beauty.and.FashionPhotographer]

Contolnet lineart seesm to only work for sd1.5 and not for sdxl, right?

[BadNewsBerrington]

Is it possible to create a closeup anime image of a face (a headshot type of image) and use ADetailer to create a full body image with that face? Hope that makes sense. I'm trying to create consistent anime characters without creating big style sheets. I would essentially always use that headshot as the face and swap it onto the new body. the other face swappers don't work with anime/cartoon style images very well and in most cases not at all.