SD.Next

SD.Next is a self-hosted web interface for artists and researchers who want to create and edit AI images and videos on their own hardware. It supports Stable Diffusion and other diffusion models.

• Image and video generation: Create images or videos from text prompts or source images, using models from the built-in download selection or models you supply.
• Editing and guidance: Revise images with inpainting, outpainting and detail refinement, and guide generation with LoRA, ControlNet and IPAdapters.
• Image and video processing: Upscale images, interpolate frames, and apply color correction, grading and filters.
• Captions and tags: Generate captions with built-in LLM and vision-language models, and tag images with WaifuDiffusion and DeepDanbooru.
• Lower GPU memory use: SDNQ supports pre-quantized models and quantization during use, while Balanced Offload shares memory use between the CPU and GPU to fit larger models on limited hardware.

SD.Next is open source under Apache 2.0. It runs on Windows and Linux with NVIDIA, AMD and Intel GPU support, and on macOS with Apple M1/M2 acceleration. CPU-only execution is also supported. Docker recipes cover CUDA, ROCm, Intel IPEX and OpenVINO.

The browser interface has desktop and mobile layouts, multiple languages and selectable themes. Custom scripts and extensions can add workflows.