With just one api function, img2img takes a new image and fixes the face right away

[joseph16388]

I wanted to use the adetailer API in img2img to repair images directly as they are generated, so I used the following request dictionary:

novel_dict = {
"prompt": tag,
"negative_prompt": negative_prompt,
"sampler_name": main_app.configs.sampler_name,
"batch_size": 1,
"n_iter": 1,
"steps": main_app.configs.steps,
"cfg_scale": main_app.configs.cfg_scale,
"width": main_app.configs.width,
"height": main_app.configs.height,
"do_not_save_grid": "true",
"denoising_strength": main_app.configs.denoising_strength,
"init_images": [init_image_base64],
"alwayson_scripts": {
"ADetailer": {
"args": [
{
"ad_model": "face_yolov8n.pt"
},

]
}
},
"override_settings": {
"sd_model_checkpoint": main_app.configs.checkpoint,
"sd_vae": main_app.configs.vae
},
}

No images are produced, but when denoising_strength is 0 it is possible to produce a fixed image of the face, but there is no img2img a new image, which is not what I want. What I want is to produce a new image and fix the face at the same time, but following the novel_dict request above doesn't do that!

@DosuBot

[dosubot[bot]]

To produce a new image and fix the face at the same time using the novel_dict request, you need to ensure that the parameters are correctly set for both image generation and face inpainting. Here's an example of how you might structure the request:

import requests

# Define the URL for the ADetailer API
url = "http://your-adetailer-api-endpoint"

# Define the payload for the request
payload = {
    "prompt": tag,
    "negative_prompt": negative_prompt,
    "sampler_name": main_app.configs.sampler_name,
    "batch_size": 1,
    "n_iter": 1,
    "steps": main_app.configs.steps,
    "cfg_scale": main_app.configs.cfg_scale,
    "width": main_app.configs.width,
    "height": main_app.configs.height,
    "do_not_save_grid": "true",
    "denoising_strength": main_app.configs.denoising_strength,
    "init_images": [init_image_base64],
    "alwayson_scripts": {
        "ADetailer": {
            "args": [
                {
                    "ad_model": "face_yolov8n.pt",
                    "ad_prompt": tag,
                    "ad_negative_prompt": negative_prompt,
                    "ad_confidence": 0.5,
                    "ad_mask_min_ratio": 0.01,
                    "ad_mask_max_ratio": 1.0,
                    "ad_mask_merge_invert": "None",
                    "ad_inpaint_only_masked": True,
                    "ad_restore_face": True,
                    "ad_steps": 28,
                    "ad_cfg_scale": 7.0,
                    "ad_sampler": "DPM++ 2M Karras"
                }
            ]
        }
    },
    "override_settings": {
        "sd_model_checkpoint": main_app.configs.checkpoint,
        "sd_vae": main_app.configs.vae
    }
}

# Send the request to the ADetailer API
response = requests.post(url, json=payload)

# Check the response
if response.status_code == 200:
    result = response.json()
    # Process the result as needed
    print("Image and face fixed successfully.")
else:
    print("Failed to process the image:", response.text)

Ensure that the denoising_strength is set appropriately (not zero) to allow for both image generation and inpainting. The alwayson_scripts section should include all necessary parameters for the ADetailer extension to function correctly ^[1].

To continue talking to Dosu, mention @dosu.

[joseph16388]

To produce a new image and fix the face at the same time using the novel_dict request, you need to ensure that the parameters are correctly set for both image generation and face inpainting. Here's an example of how you might structure the request:

import requests

# Define the URL for the ADetailer API
url = "http://your-adetailer-api-endpoint"

# Define the payload for the request
payload = {
    "prompt": tag,
    "negative_prompt": negative_prompt,
    "sampler_name": main_app.configs.sampler_name,
    "batch_size": 1,
    "n_iter": 1,
    "steps": main_app.configs.steps,
    "cfg_scale": main_app.configs.cfg_scale,
    "width": main_app.configs.width,
    "height": main_app.configs.height,
    "do_not_save_grid": "true",
    "denoising_strength": main_app.configs.denoising_strength,
    "init_images": [init_image_base64],
    "alwayson_scripts": {
        "ADetailer": {
            "args": [
                {
                    "ad_model": "face_yolov8n.pt",
                    "ad_prompt": tag,
                    "ad_negative_prompt": negative_prompt,
                    "ad_confidence": 0.5,
                    "ad_mask_min_ratio": 0.01,
                    "ad_mask_max_ratio": 1.0,
                    "ad_mask_merge_invert": "None",
                    "ad_inpaint_only_masked": True,
                    "ad_restore_face": True,
                    "ad_steps": 28,
                    "ad_cfg_scale": 7.0,
                    "ad_sampler": "DPM++ 2M Karras"
                }
            ]
        }
    },
    "override_settings": {
        "sd_model_checkpoint": main_app.configs.checkpoint,
        "sd_vae": main_app.configs.vae
    }
}

# Send the request to the ADetailer API
response = requests.post(url, json=payload)

# Check the response
if response.status_code == 200:
    result = response.json()
    # Process the result as needed
    print("Image and face fixed successfully.")
else:
    print("Failed to process the image:", response.text)

Ensure that the denoising_strength is set appropriately (not zero) to allow for both image generation and inpainting. The alwayson_scripts section should include all necessary parameters for the ADetailer extension to function correctly [1].

To continue talking to Dosu, mention @dosu.

@dosu

But when I use
novel_dict = {
"prompt": tag,
"negative_prompt": negative_prompt,
"sampler_name": main_app.configs.sampler_name,
"batch_size": 1,
"n_iter": 1,
"steps": main_app.configs.steps,
"cfg_scale": main_app.configs.cfg_scale,
"width": main_app.configs.width,
"height": main_app.configs.height,
"do_not_save_grid": "true",
"denoising_strength": 0.0,
"init_images": [init_image_base64],

"alwayson_scripts": {
"ADetailer": {
"args": [
{
"ad_model": "face_yolov8n.pt"
},

]
}
},
"override_settings": {
"sd_model_checkpoint": main_app.configs.checkpoint,
"sd_vae": main_app.configs.vae
},
}

It is possible to fix the face, but the picture has not changed except the face, which is not what I want, but it does produce the picture, and does not require too complicated ADetailer parameters!However, by modifying "denoising_strength" to be non-zero, the graph cannot be generated.

[dosubot[bot]]

To produce a new image and fix the face simultaneously without requiring too many complicated parameters, you need to ensure that the denoising_strength is set appropriately (not zero) to allow for both image generation and inpainting. Here is a simplified version of your novel_dict request:

novel_dict = {
    "prompt": tag,
    "negative_prompt": negative_prompt,
    "sampler_name": main_app.configs.sampler_name,
    "batch_size": 1,
    "n_iter": 1,
    "steps": main_app.configs.steps,
    "cfg_scale": main_app.configs.cfg_scale,
    "width": main_app.configs.width,
    "height": main_app.configs.height,
    "do_not_save_grid": "true",
    "denoising_strength": main_app.configs.denoising_strength,  # Ensure this is not zero
    "init_images": [init_image_base64],
    "alwayson_scripts": {
        "ADetailer": {
            "args": [
                {
                    "ad_model": "face_yolov8n.pt",
                    "ad_restore_face": True,  # Enable face restoration
                    "ad_inpaint_only_masked": True  # Inpaint only masked areas
                }
            ]
        }
    },
    "override_settings": {
        "sd_model_checkpoint": main_app.configs.checkpoint,
        "sd_vae": main_app.configs.vae
    }
}

Make sure the denoising_strength is set to a value that allows for both image generation and inpainting. This configuration should help you achieve the desired result of generating a new image and fixing the face at the same time ^[1].

To continue talking to Dosu, mention @dosu.