Implementation:AUTOMATIC1111 Stable diffusion webui Img2img

Knowledge Sources	stable-diffusion-webui
Domains	Image Generation, Image Editing, Inpainting, Diffusion Models
Last Updated	2026-02-08 00:00 GMT

Overview

Concrete tool for dispatching image-to-image generation requests across multiple input modes (standard, sketch, inpaint, batch) provided by the AUTOMATIC1111 stable-diffusion-webui repository.

Description

The img2img function is the top-level entry point for all image-to-image generation in the WebUI. It receives the Gradio UI state, interprets the integer mode parameter to select the appropriate image source and mask, constructs a StableDiffusionProcessingImg2Img object, and delegates to either batch processing or single-image generation.

The mode dispatch logic maps each integer to a specific image/mask extraction strategy:

• Mode 0 (img2img): Uses init_img directly, no mask.
• Mode 1 (sketch): Uses the sketch canvas input, no mask.
• Mode 2 (inpaint): Extracts image and mask from init_img_with_mask dict, converts mask to binary via create_binary_mask().
• Mode 3 (inpaint sketch): Computes mask by detecting pixel differences between the edited inpaint_color_sketch and the original inpaint_color_sketch_orig. Applies brightness adjustment with mask_alpha and Gaussian blur compositing.
• Mode 4 (inpaint upload mask): Uses separately uploaded init_img_inpaint and init_mask_inpaint.
• Mode 5 (batch): Delegates to process_batch() for directory-based or upload-based batch processing.

After mode dispatch, the function validates denoising strength is within [0.0, 1.0], optionally scales dimensions by scale_by, creates the processing object, and invokes the generation pipeline.

Usage

This function is called by the Gradio UI when the user clicks "Generate" on the img2img tab. It is also accessible through the WebUI API for programmatic image-to-image workflows.

Code Reference

Source Location

• Repository: stable-diffusion-webui
• File: modules/img2img.py
• Lines: 152-253

Signature

def img2img(
    id_task: str,
    request: gr.Request,
    mode: int,
    prompt: str,
    negative_prompt: str,
    prompt_styles,
    init_img,
    sketch,
    init_img_with_mask,
    inpaint_color_sketch,
    inpaint_color_sketch_orig,
    init_img_inpaint,
    init_mask_inpaint,
    mask_blur: int,
    mask_alpha: float,
    inpainting_fill: int,
    n_iter: int,
    batch_size: int,
    cfg_scale: float,
    image_cfg_scale: float,
    denoising_strength: float,
    selected_scale_tab: int,
    height: int,
    width: int,
    scale_by: float,
    resize_mode: int,
    inpaint_full_res: bool,
    inpaint_full_res_padding: int,
    inpainting_mask_invert: int,
    img2img_batch_input_dir: str,
    img2img_batch_output_dir: str,
    img2img_batch_inpaint_mask_dir: str,
    override_settings_texts,
    img2img_batch_use_png_info: bool,
    img2img_batch_png_info_props: list,
    img2img_batch_png_info_dir: str,
    img2img_batch_source_type: str,
    img2img_batch_upload: list,
    *args
) -> tuple[list, str, str, str]:

Import

from modules.img2img import img2img

I/O Contract

Inputs

Name	Type	Required	Description
id_task	str	Yes	Unique task identifier for progress tracking
request	gr.Request	Yes	Gradio request object containing user info
mode	int	Yes	Input mode: 0=img2img, 1=sketch, 2=inpaint, 3=inpaint sketch, 4=inpaint upload mask, 5=batch
prompt	str	Yes	Text prompt describing the desired output
negative_prompt	str	Yes	Text prompt describing undesired content
prompt_styles	list	No	Named style presets to apply
init_img	PIL.Image	No	Source image for mode 0
sketch	PIL.Image	No	Sketch canvas image for mode 1
init_img_with_mask	dict	No	Dict with "image" and "mask" keys for mode 2
inpaint_color_sketch	PIL.Image	No	User-edited image for mode 3
inpaint_color_sketch_orig	PIL.Image	No	Original unedited image for mode 3 diff computation
init_img_inpaint	PIL.Image	No	Source image for mode 4
init_mask_inpaint	PIL.Image	No	Uploaded mask for mode 4
mask_blur	int	Yes	Gaussian blur radius applied to inpainting mask
mask_alpha	float	Yes	Mask transparency for inpaint sketch mode (0-100)
inpainting_fill	int	Yes	Fill method for masked area: 0=fill, 1=original, 2=latent noise, 3=latent nothing
denoising_strength	float	Yes	Controls noise level added to input (0.0-1.0)
resize_mode	int	Yes	How to resize input: 0=just resize, 1=crop and resize, 2=resize and fill, 3=latent upscale
cfg_scale	float	Yes	Classifier-free guidance scale
width	int	Yes	Target output width in pixels
height	int	Yes	Target output height in pixels

Outputs

Name	Type	Description
images	list[PIL.Image]	Generated output images
generation_info_js	str	JSON string containing generation metadata
info_html	str	HTML-formatted generation info for UI display
comments_html	str	HTML-formatted comments and warnings

Usage Examples

Basic Usage

from modules.img2img import img2img

# Standard img2img call (mode 0)
images, info_js, info_html, comments = img2img(
    id_task="task_001",
    request=gradio_request,
    mode=0,                    # standard img2img
    prompt="a beautiful landscape painting",
    negative_prompt="blurry, low quality",
    prompt_styles=[],
    init_img=source_image,     # PIL.Image
    sketch=None,
    init_img_with_mask=None,
    inpaint_color_sketch=None,
    inpaint_color_sketch_orig=None,
    init_img_inpaint=None,
    init_mask_inpaint=None,
    mask_blur=4,
    mask_alpha=0.0,
    inpainting_fill=0,
    n_iter=1,
    batch_size=1,
    cfg_scale=7.0,
    image_cfg_scale=1.5,
    denoising_strength=0.75,
    selected_scale_tab=0,
    height=512,
    width=512,
    scale_by=1.0,
    resize_mode=0,
    inpaint_full_res=False,
    inpaint_full_res_padding=0,
    inpainting_mask_invert=0,
    img2img_batch_input_dir="",
    img2img_batch_output_dir="",
    img2img_batch_inpaint_mask_dir="",
    override_settings_texts=[],
    img2img_batch_use_png_info=False,
    img2img_batch_png_info_props=[],
    img2img_batch_png_info_dir="",
    img2img_batch_source_type="from dir",
    img2img_batch_upload=[],
)

Related Pages

Implements Principle

• Principle:AUTOMATIC1111_Stable_diffusion_webui_Input_image_mode_configuration