bing-su / adetailer Goto Github PK

awesome extension :) thank you.

Would be great to be able to do multiple passes, and choose which model each pass is utilizing.

Could probably do this by just duplicating the extension, and picking different models for each instance of the extension, but would be nicer to have built in

When using alwayson_scripts controlnet and After Detailer together, the API does not return the ControlNet processed image. The final image returns, however the canny/depth/pose reference image is not returned.

EDIT: I might have answered my own question, I see additional settings in the A1111 Settings:

• ad_save_previews
• ad_save_images_before

I'll see if these change the expected behavior.

EDIT 2: Nope, unfortunately the API still only returns the final image.

Would applying face detailer and hand detailer at the same time possible in one go using automatic1111?

Can we have the power to use a custom steps count for the adetailer pass instead of directly inheriting from the original image pass?

Would be really great to be able to store the pos and neg prompts for both 1st&2nd passes in a text file after each generate acrion, and have a button in ADetailer to restore last used prompts from previous sessions in A1111

especially now with the cool new multiple passes, it's 4 prompts fields to manually fenagle back into place each new session

as writen in the title.
1st tab full body
2nd tab face
(3rd tab hand)

This extension is amazing. I think the only thing left would be a way to do just background. The main prompt can be the main subjects, then adetailer for face, hands and hopefully background in the future for really smooth workflow.

Would be awesome to be able to tick a box and have intermediate versions saved, so a before adetailer, after 1st pass and after 2nd pass, to make comparison of each step easier when tweaking params

there is an error when i install it,i want to know how to do with it :
Error loading script: !adetailer.py
Traceback (most recent call last):
File "/root/autodl-tmp/stable-diffusion-webui/modules/scripts.py", line 256, in load_scripts
script_module = script_loading.load_module(scriptfile.path)
File "/root/autodl-tmp/stable-diffusion-webui/modules/script_loading.py", line 11, in load_module
module_spec.loader.exec_module(module)
File "", line 848, in exec_module
File "", line 219, in _call_with_frames_removed
File "/root/autodl-tmp/stable-diffusion-webui/extensions/adetailer-main/scripts/!adetailer.py", line 15, in
from adetailer import (
File "/root/autodl-tmp/stable-diffusion-webui/extensions/adetailer-main/adetailer/init.py", line 5, in
from .ultralytics import ultralytics_predict
File "/root/autodl-tmp/stable-diffusion-webui/extensions/adetailer-main/adetailer/ultralytics.py", line 59, in
def mask_to_pil(masks, shape: tuple[int, int]) -> list[Image.Image]:
TypeError: 'type' object is not subscriptable

for example, i have a images shows a female, and there are a lot of in the background, how to prevent adetailer detailing those people's face? it will make picture looks weird. such as the picture below

항상 enable 이 체크되어있습니다.

다른 익스텐션들은 사용할때 켜야 하는데

ad 는 사용안할때 꺼야 하는데... 의도된 건가요?

예전이라면 항상 켜져있어도 모델이 None 으로 되어있어 적용이 안되었는데

디폴트를 끔으로 하면 어떨까요

When I choose to use "hand_yolov8n.pt", the following result appears:

PS: All parameters in [adetailer] are using default values, I don't know where this error comes from...😣

Great addon! I am wondering is the A1111 API would support this?

I read these docs and it got me excited that it might be possible?

https://github.com/mix1009/sdwebuiapi#scripts-support

Any hints you could give would be super awesome!!!

업데이트 후 혹시 ARGS 바뀐 부분이 있나요? 기존 방식으로 API 호출하면 또 에러가 나서 롤백 하였습니다.

현재 API 로 호출하는 파라메터는 이렇습니다.

  alwayson_scripts = {
    "After Detailer":
      {
        "args": [
          true,
          args[1]["ad_model"],
          args[1]["ad_prompt"],
          args[1]["ad_negative_prompt"],
          args[1]["ad_conf"].to_i,
          args[1]["ad_dilate_erode"].to_i,
          args[1]["ad_x_offset"].to_i,
          args[1]["ad_y_offset"].to_i,
          args[1]["ad_mask_blur"].to_i,
          args[1]["ad_denoising_strength"].to_f,
          args[1]["ad_inpaint_full_res"],
          args[1]["ad_inpaint_full_res_padding"].to_i,
          args[1]["ad_use_inpaint_width_height"],
          args[1]["ad_inpaint_width"].to_i,
          args[1]["ad_inpaint_height"].to_i,
          args[1]["ad_cfg_scale"].to_f,
          "None",
          1,
        ]
      }
  }
  
  8c525a3c 커밋에서는 정상동작 합니다. 혹시 추가된 부분이 있으면 귀뜸해주시면 수정해보겠습니다.

에러 메세지 첨부 합니다.

May 10 14:38:34 rubyon.co.kr sh[100987]: Error running process: /home/ai/stable-diffusion-webui/extensions/adetailer/scripts/!adetailer.py
May 10 14:38:34 rubyon.co.kr sh[100987]: Traceback (most recent call last):
May 10 14:38:34 rubyon.co.kr sh[100987]: File "/home/ai/stable-diffusion-webui/extensions/adetailer/scripts/!adetailer.py", line 346, in get_args
May 10 14:38:34 rubyon.co.kr sh[100987]: inp = get_one_args(enabled, *args)
May 10 14:38:34 rubyon.co.kr sh[100987]: File "/home/ai/stable-diffusion-webui/extensions/adetailer/adetailer/args.py", line 117, in get_one_args
May 10 14:38:34 rubyon.co.kr sh[100987]: return ADetailerArgs(**arg_dict)
May 10 14:38:34 rubyon.co.kr sh[100987]: File "pydantic/main.py", line 341, in pydantic.main.BaseModel.init
May 10 14:38:34 rubyon.co.kr sh[100987]: pydantic.error_wrappers.ValidationError: 2 validation errors for ADetailerArgs
May 10 14:38:34 rubyon.co.kr sh[100987]: ad_use_cfg_scale
May 10 14:38:34 rubyon.co.kr sh[100987]: value could not be parsed to a boolean (type=type_error.bool)
May 10 14:38:34 rubyon.co.kr sh[100987]: ad_cfg_scale
May 10 14:38:34 rubyon.co.kr sh[100987]: value is not a valid float (type=type_error.float)
May 10 14:38:34 rubyon.co.kr sh[100987]: The above exception was the direct cause of the following exception:
May 10 14:38:34 rubyon.co.kr sh[100987]: Traceback (most recent call last):
May 10 14:38:34 rubyon.co.kr sh[100987]: File "/home/ai/stable-diffusion-webui/modules/scripts.py", line 417, in process
May 10 14:38:34 rubyon.co.kr sh[100987]: script.process(p, *script_args)
May 10 14:38:34 rubyon.co.kr sh[100987]: File "/home/ai/stable-diffusion-webui/extensions/adetailer/scripts/!adetailer.py", line 558, in process
May 10 14:38:34 rubyon.co.kr sh[100987]: arg_list = self.get_args(*args_)
May 10 14:38:34 rubyon.co.kr sh[100987]: File "/home/ai/stable-diffusion-webui/extensions/adetailer/scripts/!adetailer.py", line 355, in get_args
May 10 14:38:34 rubyon.co.kr sh[100987]: raise ValueError("\n".join(message)) from e
May 10 14:38:34 rubyon.co.kr sh[100987]: ValueError: [-] ADetailer: ValidationError when validating 1st arguments: 2 validation errors for ADetailerArgs
May 10 14:38:34 rubyon.co.kr sh[100987]: ad_use_cfg_scale
May 10 14:38:34 rubyon.co.kr sh[100987]: value could not be parsed to a boolean (type=type_error.bool)
May 10 14:38:34 rubyon.co.kr sh[100987]: ad_cfg_scale
May 10 14:38:34 rubyon.co.kr sh[100987]: value is not a valid float (type=type_error.float)
May 10 14:38:34 rubyon.co.kr sh[100987]: ad_model: 'face_yolov8s.pt' (<class 'str'>)
May 10 14:38:34 rubyon.co.kr sh[100987]: ad_prompt: '' (<class 'str'>)
May 10 14:38:34 rubyon.co.kr sh[100987]: ad_negative_prompt: '' (<class 'str'>)
May 10 14:38:34 rubyon.co.kr sh[100987]: ad_conf: 30 (<class 'int'>)
May 10 14:38:34 rubyon.co.kr sh[100987]: ad_dilate_erode: 32 (<class 'int'>)
May 10 14:38:34 rubyon.co.kr sh[100987]: ad_x_offset: 0 (<class 'int'>)
May 10 14:38:34 rubyon.co.kr sh[100987]: ad_y_offset: 0 (<class 'int'>)
May 10 14:38:34 rubyon.co.kr sh[100987]: ad_mask_blur: 4 (<class 'int'>)
May 10 14:38:34 rubyon.co.kr sh[100987]: ad_denoising_strength: 0.4 (<class 'float'>)
May 10 14:38:34 rubyon.co.kr sh[100987]: ad_inpaint_full_res: True (<class 'bool'>)
May 10 14:38:34 rubyon.co.kr sh[100987]: ad_inpaint_full_res_padding: 0 (<class 'int'>)
May 10 14:38:34 rubyon.co.kr sh[100987]: ad_use_inpaint_width_height: False (<class 'bool'>)
May 10 14:38:34 rubyon.co.kr sh[100987]: ad_inpaint_width: 512 (<class 'int'>)
May 10 14:38:34 rubyon.co.kr sh[100987]: ad_inpaint_height: 512 (<class 'int'>)
May 10 14:38:34 rubyon.co.kr sh[100987]: ad_use_cfg_scale: 7.0 (<class 'float'>)
May 10 14:38:34 rubyon.co.kr sh[100987]: ad_cfg_scale: 'None' (<class 'str'>)
May 10 14:38:34 rubyon.co.kr sh[100987]: ad_controlnet_model: 1 (<class 'int'>)
May 10 14:38:34 rubyon.co.kr sh[100987]: ad_controlnet_weight: 1.0 (<class 'float'>)

Error loading script: !adetailer.py Traceback (most recent call last): File "/home/*/stable-diffusion-webui/modules/scripts.py", line 256, in load_scripts script_module = script_loading.load_module(scriptfile.path) File "/home/*/stable-diffusion-webui/modules/script_loading.py", line 11, in load_module module_spec.loader.exec_module(module) File "<frozen importlib._bootstrap_external>", line 848, in exec_module File "<frozen importlib._bootstrap>", line 219, in _call_with_frames_removed File "/home/j*/stable-diffusion-webui/extensions/adetailer/scripts/!adetailer.py", line 17, in <module> from adetailer import ( File "/home/*/stable-diffusion-webui/extensions/adetailer/adetailer/__init__.py", line 2, in <module> from .args import AD_ENABLE, ALL_ARGS, ADetailerArgs, enable_check File "/home/*/stable-diffusion-webui/extensions/adetailer/adetailer/args.py", line 23, in <module> class ArgsList(UserList): File "/home/*/stable-diffusion-webui/extensions/adetailer/adetailer/args.py", line 25, in ArgsList def attrs(self) -> tuple[str]: TypeError: 'type' object is not subscriptable
i have the latest AUTOMATIC1111/stable-diffusion-webui
The latest adetailer a0cd6f1 (Sat May 13 06:23:56 2023)
Linux Mint OS

Yolo dosn't work, but i can use mediapipe.
Progress stops at 98%, and the progress in the interface just freezes.

I also have ultralytics

But then again, mediapipe works as it should - it runs at the end of generation and improves the picture.
Whats wrong?

Feature request to support optional usage of the ControlNet inpainting model when this extension is enabled, as I believe this could improve results, especially at higher denoise values. More info here: Mikubill/sd-webui-controlnet#968

This could be handled with the external code API: https://github.com/Mikubill/sd-webui-controlnet/wiki/API#external-code-api

When I use openpose and Adetailer inpaints the face, I notice I try to use openpose's full frame, instead of just the face portion.
Resulting in the faces regularly having tiny bodies inside.
Because it is using the full scale openpose model, instead of it's face portion.

Love this extension, hopefully it's something that can be fixed in the future.

Thanks for creating and sharing it!

I have previews enabled, showing progress for every step (0).
But during the adetailer pass at the end of regular generation, previews are completely frozen until the very last step.

I'm using the current commit (d95d82c) with the current (5ab7f213) version of sd-webui (torch2, xformers 0.0.17, video card is a 1070 and running ubuntu 22.10).

There are no errors in stdout and everything else functions properly.

Error running postprocess_batch: F:\AI\stable-diffusion-webui\extensions\adetailer\scripts!adetailer.py00:00<?, ?it/s]
Traceback (most recent call last):
File "F:\AI\stable-diffusion-webui\modules\scripts.py", line 463, in postprocess_image
script.postprocess_image(p, pp, *script_args)
File "F:\AI\stable-diffusion-webui\extensions\adetailer\scripts!adetailer.py", line 615, in postprocess_image
is_processed |= self._postprocess_image(p, pp, args, n=n)
File "F:\AI\stable-diffusion-webui\extensions\adetailer\scripts!adetailer.py", line 586, in _postprocess_image
processed = process_images(p2)
File "F:\AI\stable-diffusion-webui\modules\processing.py", line 526, in process_images
res = process_images_inner(p)
File "F:\AI\stable-diffusion-webui\extensions\sd-webui-controlnet\scripts\batch_hijack.py", line 42, in processing_process_images_hijack
return getattr(processing, '__controlnet_original_process_images_inner')(p, *args, **kwargs)
File "F:\AI\stable-diffusion-webui\modules\processing.py", line 615, in process_images_inner
p.init(p.all_prompts, p.all_seeds, p.all_subseeds)
File "F:\AI\stable-diffusion-webui\modules\processing.py", line 1065, in init
image_masked.paste(image.convert("RGBA").convert("RGBa"), mask=ImageOps.invert(self.mask_for_overlay.convert('L')))
File "F:\AI\stable-diffusion-webui\venv\lib\site-packages\PIL\Image.py", line 1731, in paste
self.im.paste(im, box, mask.im)
ValueError: images do not match

any ideas?

with ddetailer I could send image to inpaint, and press only masked and set width/height to 768/768 for example, and it would inpaint the faces at that resolution, when doing the same with adetailer I get an error, is there a different way to do it with adetailer that I'm missing?

Arguments: ('task(2rp4l4pchnhlx65)', 2, 'absurdres, highres, best quality, 1girl, ', '(worst quality:2), (low quality:2), (normal quality:2), lowres, jpeg artifacts, signature, watermark, username, blurry, artist name, text, error', [], <PIL.Image.Image image mode=RGBA size=1024x1536 at 0x1ADA5407BB0>, None, {'image': <PIL.Image.Image image mode=RGBA size=1024x1536 at 0x1ADA54074F0>, 'mask': <PIL.PngImagePlugin.PngImageFile image mode=RGBA size=1024x1536 at 0x1ADA54075B0>}, None, None, None, None, 20, 15, 4, 0, 1, False, False, 1, 1, 7, 1.5, 0.5, -1.0, -1.0, 0, 0, 0, False, 0, 1536, 1024, 1, 0, 1, 32, 0, '', '', '', [], 0, True, 'None', '', '', 30, 32, 0, 0, 4, 0.4, True, 0, True, 528, 1024, 7, 'control_v11p_sd15_inpaint [ebff9138]', 1, False, 'MultiDiffusion', False, 10, 10, 1, 64, False, True, 1024, 1024, 96, 96, 48, 1, 'None', 2, False, False, False, False, False, 0.4, 0.4, 0.2, 0.2, '', '', 'Background', 0.2, False, 0.4, 0.4, 0.2, 0.2, '', '', 'Background', 0.2, False, 0.4, 0.4, 0.2, 0.2, '', '', 'Background', 0.2, False, 0.4, 0.4, 0.2, 0.2, '', '', 'Background', 0.2, False, 0.4, 0.4, 0.2, 0.2, '', '', 'Background', 0.2, False, 0.4, 0.4, 0.2, 0.2, '', '', 'Background', 0.2, False, 0.4, 0.4, 0.2, 0.2, '', '', 'Background', 0.2, False, 0.4, 0.4, 0.2, 0.2, '', '', 'Background', 0.2, False, False, True, True, False, 3072, 192, True, False, 1, False, False, False, 1.1, 1.5, 100, 0.7, False, False, True, False, False, 0, 'Gustavosta/MagicPrompt-Stable-Diffusion', '', False, 7, 100, 'Constant', 0, 'Constant', 0, 4, False, False, 'LoRA', 'None', 1, 1, 'LoRA', 'None', 1, 1, 'LoRA', 'None', 1, 1, 'LoRA', 'None', 1, 1, 'LoRA', 'None', 1, 1, None, 'Refresh models', <controlnet.py.UiControlNetUnit object at 0x000001AC28652980>, 'NONE:0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0\nALL:1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1\nINS:1,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0,0\nIND:1,0,0,0,1,1,1,0,0,0,0,0,0,0,0,0,0\nINALL:1,1,1,1,1,1,1,0,0,0,0,0,0,0,0,0,0\nMIDD:1,0,0,0,1,1,1,1,1,1,1,1,0,0,0,0,0\nOUTD:1,0,0,0,0,0,0,0,1,1,1,1,0,0,0,0,0\nOUTS:1,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1\nOUTALL:1,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,1\nALL0.5:0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5', True, 0, 'values', '0,0.25,0.5,0.75,1', 'Block ID', 'IN05-OUT05', 'none', '', '0.5,1', 'BASE,IN00,IN01,IN02,IN03,IN04,IN05,IN06,IN07,IN08,IN09,IN10,IN11,M00,OUT00,OUT01,OUT02,OUT03,OUT04,OUT05,OUT06,OUT07,OUT08,OUT09,OUT10,OUT11', 1.0, 'black', '20', False, 'ATTNDEEPON:IN05-OUT05:attn:1\n\nATTNDEEPOFF:IN05-OUT05:attn:0\n\nPROJDEEPOFF:IN05-OUT05:proj:0\n\nXYZ:::1', False, False, False, False, False, '1:1,1:2,1:2', '0:0,0:0,0:1', '0.2,0.8,0.8', 20, '<ul>\n<li><code>CFG Scale</code> should be 2 or lower.</li>\n</ul>\n', True, True, '', '', True, 50, True, 1, 0, False, 4, 0.5, 'Linear', 'None', '<p style="margin-bottom:0.75em">Recommended settings: Sampling Steps: 80-100, Sampler: Euler a, Denoising strength: 0.8</p>', 128, 8, ['left', 'right', 'up', 'down'], 1, 0.05, 128, 4, 0, ['left', 'right', 'up', 'down'], False, False, 'positive', 'comma', 0, False, False, '', '<p style="margin-bottom:0.75em">Will upscale the image by the selected scale factor; use width and height sliders to set tile size</p>', 64, 0, 2, 1, '', [], 0, '', [], 0, '', [], True, False, False, False, 0, None, False, 50) {}
Traceback (most recent call last):
  File "C:\Users\apyr\Documents\auto1111\stable-diffusion-webui\modules\call_queue.py", line 57, in f
    res = list(func(*args, **kwargs))
  File "C:\Users\apyr\Documents\auto1111\stable-diffusion-webui\modules\call_queue.py", line 37, in f
    res = func(*args, **kwargs)
  File "C:\Users\apyr\Documents\auto1111\stable-diffusion-webui\modules\img2img.py", line 181, in img2img
    processed = process_images(p)
  File "C:\Users\apyr\Documents\auto1111\stable-diffusion-webui\modules\processing.py", line 515, in process_images
    res = process_images_inner(p)
  File "C:\Users\apyr\Documents\auto1111\stable-diffusion-webui\extensions\sd-webui-controlnet\scripts\batch_hijack.py", line 42, in processing_process_images_hijack
    return getattr(processing, '__controlnet_original_process_images_inner')(p, *args, **kwargs)
  File "C:\Users\apyr\Documents\auto1111\stable-diffusion-webui\modules\processing.py", line 604, in process_images_inner
    p.init(p.all_prompts, p.all_seeds, p.all_subseeds)
  File "C:\Users\apyr\Documents\auto1111\stable-diffusion-webui\modules\processing.py", line 1014, in init
    mask = mask.crop(crop_region)
  File "C:\Users\apyr\Documents\auto1111\stable-diffusion-webui\venv\lib\site-packages\PIL\Image.py", line 1228, in crop
    raise ValueError(msg)
ValueError: Coordinate 'right' is less than 'left'

it always "None", how to setup the Control net properly model?

Just a question, but does the prompt during the AD pass take into account TI embedding and LORA placeholder texts?

This is a great plugin.
I followed the instructions in args.py file for setup, but for some reason, I'm encountering an error.

I installed the git through A1111's extension tab, click reload, the CMD output hung at "Restarting WebUI". I left it for a few hours before forcing it close.

When running webui-user.bat it hangs at:
warnings.warn(

I deleted the extension folder then ran it again and I'm back to normal. I noticed the next line for me would normally be:
ControlNet v1.1.133

So maybe there's a conflict with ControlNet?

I have the latest version of SD and Adetailer.
I can't get it to run properly.
I restarted the webUI, restarted SD.
I enabled both settings for saving copies of mask and before-Adetailer.

I am using the default settings of SD.
SD Prompt: person (doesn't matter what I choose)
Adetailer prompt: face
Model: 8n (doesn't matter which I choose)
Adetailer enabled, doesn't matter what other settings (I have also tried checking Use Inpaint Width/height).
I get to 95%:

In my console I have this (typically the middle bar shows as 95%, same as WebUI image above):

When I Ctrl + C to kill it, I get:
Interrupted with signal 2 in <frame at 0x000001FFA0CE89B0, file 'c:\\webui\\webui.py', line 266, code wait_on_server>

After updating to Torch 2, i did a fresh install of SD.... in img2img i use Controlnet Tile + Ultimate SD Upscaler (4xultrasharp)....and aDetailer.
when i turn on ADETAILER, the faces are a garbled mess now.... when i turn it off.... it's working same as it was before the Torch update ....
is there some incompatibility or change that you know of that would cause this?

I downloaded face_yolov8n_v2.pt but it didn't show up in the model list.

There is currently a huge downside for adetailer, and that is we cannot use it on images with multiple subjects.
For example if we have 5 people in an image, adetailer face will modify all subjects.

If we can have an option to only target subject at index N (of the list of detected targets), this will greatly enhance the usability of adetailer.
For example if our image has a man and a woman, we can limit 1st tab to index 0 and inpaint a man, and limit 2nd tab to index 1 to inpaint the woman.

I have another extension, similar to adetailer, however when trying to do comparison runs, disabling adetailer (latest version), its causing issues anyway

Error running postprocess_batch: X:\STABLEDIFFUSION\AUTOMATIC111\extensions\adetailer\scripts!adetailer.py
Traceback (most recent call last):
File "X:\STABLEDIFFUSION\AUTOMATIC111\modules\scripts.py", line 462, in postprocess_image
script.postprocess_image(p, pp, *script_args)
File "X:\STABLEDIFFUSION\AUTOMATIC111\extensions\adetailer\scripts!adetailer.py", line 438, in postprocess_image
args = get_args(*args_)
File "X:\STABLEDIFFUSION\AUTOMATIC111\extensions\adetailer\adetailer\args.py", line 73, in get_args
arg_dict = {all_args.attr: args[i] for i, all_args in enumerate(ALL_ARGS)}
File "X:\STABLEDIFFUSION\AUTOMATIC111\extensions\adetailer\adetailer\args.py", line 73, in
arg_dict = {all_args.attr: args[i] for i, all_args in enumerate(ALL_ARGS)}
IndexError: tuple index out of range

This causes automatic to become unresponsive unless closed

In windows, models are downloaded into C:\Users\Administrator.cache\huggingface\hub\models--Bingsu--adetailer\snapshots\403e56c1b291da2145da824f3fde640bbe08de2f with shotcut to C:\Users\Administrator.cache\huggingface\hub\models--Bingsu--adetailer\blobs
Any way to use ./models/adetailer directly? It is bad network to download files from huggingface

I have this installed, but I have no idea how to use it properly. Can you explain?

a0cd6f1

看了后台没有显示错误

Hi. Thank you for the very cool extension, it helps just unreal!

But I don't fully understand what each slider is responsible for.

Can you please describe each function briefly and for the dumb ones. A couple of words would be enough. Or, even better, release an update where this explanation pops up when you point your cursor?
I don't really understand neural network theory, but I wish I understood what I was poking at. XD
If it's not too much trouble for you, of course.
(Sorry for the bad English)

Error running postprocess_batch: /root/stable-diffusion-webui/extensions/adetailer/scripts/!adetailer.py
Traceback (most recent call last):
File "/root/stable-diffusion-webui/modules/scripts.py", line 462, in postprocess_image
script.postprocess_image(p, pp, *script_args)
File "/root/stable-diffusion-webui/extensions/adetailer/scripts/!adetailer.py", line 642, in postprocess_image
is_processed |= self._postprocess_image(p, pp, args, n=n)
File "/root/stable-diffusion-webui/extensions/adetailer/scripts/!adetailer.py", line 587, in _postprocess_image
pred = predictor(ad_model, pp.image, args.ad_conf, **kwargs)
File "/root/stable-diffusion-webui/extensions/adetailer/adetailer/ultralytics.py", line 25, in ultralytics_predict
from ultralytics import YOLO
ModuleNotFoundError: No module named 'ultralytics'

Basically what the title says, I want to generate a completely new generation with completely different features but I want a face from existing images I have to this new generation.

Side note I suggest you enable the discussions tab since this question doesn't really belong in issues

I have put face_yolo8s.pt into ./models/adetailer/face_yolov8s.pt, but when i enable adetailer, it said ValueError: [-] ADetailer: Model 'face_yolo8s.pt' not found. Available models: ['face_yolov8n.pt', 'face_yolov8s.pt', 'mediapipe_face_full', 'mediapipe_face_short', 'hand_yolov8n.pt']

closed by restart sdwebui

Currently, the img2img process always uses the same width/height as the txt2img process:

When the Inpaint at full resolution option is enabled, this probably should instead be unlinked and be controllable by the user, as quite often the purpose of using that option in img2img normally is to allow for inpainting at a much higher resolution that differs from the source image, such as 768x768, or even beyond that, as the image gets cropped to the masked area and upscaled, and then denoised. This is especially useful if running this extension with hires fix, as we may want to inpaint at a much higher resolution than what was the resolution pre-hires fix.

Hello, Yesterday everything was working great but today when I started My webui adetailer was gone I updated it but still not showin and I get this error in my terminal window , btw I am on mac system

Error loading script: !adetailer.py

it was really great extension and I don't know what Happened !!??

As the title says, would be great to have a setting similar to 'Settings>Image Settings>Save a copy of image before doing face restoration'.

When enabled, it will save a copy before applying the face restoration.
This is good for comparisons, especially in batch processing.
I'm aware of the 'save mask previews' setting, but it would be great to also have an option for it spit out a clean image for comparison.

Thanks!

안녕하세요 프롬프츠 운영자 입니다.

마지막 커밋을 보니 디폴트 모델 값을 "None" 에서 모델 배열 첫번째 것으로 가져오는 것으로 바꾸었는데

이로인해 API 를 이용한 모든 이미지 생성시 Adetailer 가 적용되더라고요... 그래서 쭉 지켜본결과...

ㅠㅠ

Adetailer 체크박스를 해제 했음에도 불구하고 무조건 적용되는거 같아 보입니다.

기존에는 무조건 적용되더라도 기본 모델이 "None" 이어서 바이패스 된거같아 보입니다.

현재 마지막 커밋중 디폴트 모델 선택하는 부분만 급하게 수정해서 사용중인데 확인해 주셨으면 해서 글 남겨봅니다.

Receiving the following issue:

Error running postprocess_batch: C:\Users\User\Documents\stable-diffusion-webui\extensions\adetailer\scripts!adetailer.py
Traceback (most recent call last):
File "C:\Users\User\Documents\stable-diffusion-webui\modules\scripts.py", line 462, in postprocess_image
script.postprocess_image(p, pp, *script_args)
File "C:\Users\User\Documents\stable-diffusion-webui\extensions\adetailer\scripts!adetailer.py", line 446, in postprocess_image
i2i = self.get_i2i_p(p, args, pp.image)
File "C:\Users\User\Documents\stable-diffusion-webui\extensions\adetailer\scripts!adetailer.py", line 409, in get_i2i_p
self.update_controlnet_args(i2i, args)
File "C:\Users\User\Documents\stable-diffusion-webui\extensions\adetailer\scripts!adetailer.py", line 424, in update_controlnet_args
self.controlnet_ext.update_scripts_args(
File "C:\Users\User\Documents\stable-diffusion-webui\extensions\adetailer\controlnet_ext\controlnet_ext.py", line 67, in update_scripts_args
self._update_scripts_args(p, model, weight)
File "C:\Users\User\Documents\stable-diffusion-webui\extensions\adetailer\controlnet_ext\controlnet_ext.py", line 57, in _update_scripts_args
control_mode=self.external_cn.ControlMode.BALANCED,
AttributeError: module 'extensions.sd-webui-controlnet.scripts.external_code' has no attribute 'ControlMode'

Pls advice!

AD works fine with HiRes fix if "Inpaint at full resolution" is checked.

But if "Inpaint at full resolution" is unchecked, the original low res image (before hi res fix is applied) is returned as final result

AD Preview Img (1024 x 1360)

AD Before Img (1024 x 1360)

Final Img (512 x 680)

Adetailer being on by default now means that I often miss the setting on my initial generation on a webui load. Not always so bad when using txt2img on a low batch count, but not good when using img2img on an image with many people.

I've looked through the ui-config.json for an option to disable by default but cannot find it. This was not an issue when the default model was none.

I would please request restoring adetailer being disabled by default, or add a None model selection so that it isn't triggering when undesired. Additionally, can you add an easily-identifiable (e.g. adetailer.py/XXX2img/Adetailer enabled: true) setting in ui-config.json or point out which one it is currently?

使用方法：安装好插件就出现SEED下面，勾选开启，并选择模型，下面还要选“INPOINT”，还有选“CON V11p SD15”（所有模式通用）红色面部追踪框：在SD的总设置里面找到它勾选开启，这样能在生成图片的文件里面夹看到2张相同的图片，其中1张有脸部红框，否则看不到。

红色面部跟踪框，这个工具在哪儿设置？“CON V11p SD15”（所有模式通用）这个在哪儿？

我是直接在hugging下载的模型，放在哪里？stable-diffusion-webui\extensions\adetailer里，还是stable-diffusion-webui\models\adetailer里？

I tried to use this + controlnet openpose to produce batchs of images ,but i found only the first image could be controlled by controlnet model, others were random poses.
I'm using controlnet 1.125 with model sd15_openpose

"Thank you very much for your sharing. Could you please provide details on how to use it? I installed the plugin and launched it in T2IMG or IMAG2IMG, but the red facial detection box did not appear on the image. Thank you."

If I checked Hires.fix it produces this result.
I didn't adjust any parameters of this extension, just choose a adetailer model and run it.
I tried other face models and found that only the face_yolov8s.pt can work correctly.

OK,The bad news is that this happens to all models now.😂