How automatic1111/stable-diffusion-webui works
Stable Diffusion web UI is a Gradio application for generating and editing images, with an API and extensible scripts. The main user journey is entering prompts and generation settings, running txt2img or img2img, then viewing or saving images with generation metadata. The graph also preserves documented model, sampler, restoration, upscaling, training, and extension capabilities. Several subsystem paths come from the tree and README rather than inspected implementation; edges are limited to documented behavior or relationships listed in the source index.
Main entry points: processing.py Processing pipeline · ui.py Gradio UI · api.py Generation API · sd_models.py Checkpoint loading
Components 21
User interfaces
- Gradio UI
modules/ui.py - Generation API
modules/api/api.py - Image parameters
modules/infotext_utils.py
Image generation
- Text-to-image
modules/txt2img.py - Image-to-image
modules/img2img.py - Processing pipeline
modules/processing.py - Samplers
modules/sd_samplers.py - Image handling
modules/images.py
Models and conditioning
- Checkpoint loading
modules/sd_models.py - VAE loading
modules/sd_vae.py - Model conditioning
modules/sd_hijack.py - Device support
modules/devices.py
Extensions and tools
- Textual inversion
modules/textual_inversion/textual_inversion.py - Hypernetworks
modules/hypernetworks/hypernetwork.py - LoRA networks
extensions-builtin/Lora/networks.py - Extras processing
modules/extras.py - Postprocessing
modules/postprocessing.py - Embedding training
modules/textual_inversion/textual_inversion.py - Custom scripts
modules/scripts.py - Upscalers
modules/upscaler.py
Other components
- User
Connections 24
- User → Gradio UI: sets options
README.md - User → Generation API: requests generation
README.md - Processing pipeline → Samplers: uses samplers
modules/processing.py - Processing pipeline → Model conditioning: uses conditioning
modules/processing.py - Processing pipeline → Device support: uses devices
modules/processing.py - Processing pipeline → Textual inversion: applies networks
modules/processing.py - Checkpoint loading → VAE loading: loads VAE
modules/sd_models.py - Checkpoint loading → Device support: uses devices
modules/sd_models.py - Checkpoint loading → Processing pipeline: coordinates reloads
modules/sd_models.py - Model conditioning → Textual inversion: uses embeddings
modules/sd_hijack.py - Model conditioning → Hypernetworks: uses hypernetworks
modules/sd_hijack.py - Textual inversion → Image handling: uses image tools
modules/textual_inversion/textual_inversion.py - Generation API → Samplers: uses samplers
modules/api/api.py - Generation API → Image handling: uses image tools
modules/api/api.py - Generation API → Postprocessing: invokes postprocessing
modules/api/api.py - Gradio UI → Custom scripts: loads scripts
modules/ui.py - Gradio UI → Checkpoint loading: supports checkpoint selection
modules/ui.py - Gradio UI → Extras processing: offers Extras
README.md - Extras processing → Upscalers: offers upscaling
README.md - User → Embedding training: trains embeddings
README.md - Image handling → Image parameters: reads image parameters
modules/infotext_utils.py - User → Text-to-image: generates images
README.md - User → Image-to-image: edits images
README.md - User → LoRA networks: uses LoRAs
README.md