sd.cpp-webui is a simple, lightweight Gradio-based web interface for stable-diffusion.cpp.

Designed to run local AI text-to-image and video generation models without heavy dependencies, this program acts primarily as a command generator for stable-diffusion.cpp, with some extra features for image management.

• Lightweight: Minimal Python dependencies (Gradio is the main requirement, no PyTorch required).
• Feature support: Supports most stable-diffusion.cpp features (missing features are work in progress).
• Generation preview: Live image preview using stable-diffusion.cpp's native preview option.
• Image management: Built-in image gallery.
• Metadata reader: Built-in metadata reader to extract generation data from images.
• Model manager: Built-in Models tab to browse, rename and remove the installed models of each type.
• Secure access: Optional protected login with user credentials.
• Extensions: Optional plugins that add tabs, tools and model downloads (see Extensions).

This WebUI, by using stable-diffusion.cpp as its core, supports a large number of image/video generation models, including:

• SD1.x, SD2.x, SD-Turbo, SDXL, SDXL-Turbo, NitroFusion
• SD3, SD3.5, FLUX.1-Krea-dev, FLUX.1-dev, FLUX.1-schnell, FLUX.2-dev, Chroma, Qwen Image, Z-Image-Turbo
• FLUX.1-Kontext-dev, Qwen Image Edit, Qwen Image Edit 2509, Qwen Image Layered
• Wan2.1, Wan2.2 (including S2V)
• AnimateDiff (SD 1.5)
• TAESD
• PhotoMaker
• ControlNet
• ADetailer (YOLOv8 detectors)
• IP-Adapter (classic and Plus)
• PuLID-Flux
• Upscaling models
• LoRAs and embeddings

• Python (3.13 on Windows)
• Git (Recommended)

• Detailed instructions for Linux/MacOS: sdcpp_webui.sh

• Detailed instructions for Windows: sdcpp_webui_windows.ps1

• Quick Start for experienced users:

  1. Clone the repository:

     git clone https://github.com/daniandtheweb/sd.cpp-webui.git

  2. Obtain the stable-diffusion.cpp binary (sd-cli for Linux/MacOS or sd-cli.exe for Windows, plus sd-server or sd-server.exe if you want to use the --server mode) by compiling or downloading it from the releases and place it in the main sd.cpp-webui folder.

  3. Run sdcpp_webui.sh if you're on Linux/MacOS or sdcpp_webui_windows.ps1 if you're on Windows.

  4. Access the WebUI with the browser at http://localhost:7860/.

Extensions add features without touching the app. They are off by default: start the WebUI with --extensions to enable them (for example ./sdcpp_webui.sh --extensions).

Four come with the repository, in the extensions folder:

• Civitai: search Civitai and download models straight into the right model folder.
• Masker: create foreground masks from a photo with rembg and reuse them in img2img.
• Pose: turn a photo of a person into an OpenPose skeleton to use as a ControlNet control image.
• Ultimate SDCPP Upscale: upscale the images of txt2img, img2img and imgedit tile by tile, with its own model, prompts, sampler, denoise, seam fixing and ControlNet tile, and a live preview of the tiles. Based on the ideas of ComfyUI_UltimateSDUpscale and ultimate-upscale-for-automatic1111.

Only install extensions you trust. An extension is Python code that runs with all your permissions, and so are the packages it requires. The first time the WebUI sees a new or updated extension it shows what it declares (network sites, output folder, packages) in the terminal and asks for your confirmation. Declining leaves the extension unloaded and it will ask again at the next start.

To write your own, see docs/extensions.md.

Thank you to all the contributors!

• stable-diffusion.cpp - https://github.com/leejet/stable-diffusion.cpp
• Gradio - https://github.com/gradio-app/gradio