SD.Next
github.com
- Free plan
- Windows
- macOS
- Linux
- Web
- API
- Self-hosted
What SD.Next is
SD.Next is an open-source server application and web interface for generating, editing, captioning, upscaling, and processing AI images and videos. Its workflows include text-to-image, image-to-image, text-to-video, image-to-video, inpainting, and outpainting, with PNG and JPG output. The project supports many diffusion models and can automatically download selected reference models. Guidance options include LoRA, ControlNet, IPAdapters, and Prompt Enhance; processing tools cover captioning, tagging, upscaling, interpolation, colorization, filtering, and color correction. SDNQ quantization is described as reducing VRAM use by up to 4x, while Balanced Offload balances CPU and GPU memory for larger models on limited hardware. Listed hardware support includes CUDA, ROCm, ZLUDA, Intel IPEX XPU, OpenVINO, DirectML, Apple M1/M2 MPS, ONNX/Olive, and CPU-only execution. The project provides installation guidance for Linux, Mac, and Windows, as well as API access. Its security policy says it does not send or share data externally, track user or system data, store user inputs or detection results, or use analytics. The repository identifies Apache-2.0 as its license.
Who it suits
SD.Next is suited to artists, researchers, and AI enthusiasts who want a self-hosted image and video workflow. Its memory features and CPU-only option may matter to users working with limited hardware.
Good to know, and what to watch
- Open-source and licensed under Apache-2.0
- Supports image and video workflows
- Lists multiple hardware backends, including CPU-only
- SDNQ can reduce VRAM use by up to 4x
- Includes API access
- Extensions can be downloaded when enabled
- Some external connections occur for downloads and upgrades
Plans
Every plan and priceSD.Next
Open source
The verdict
SD.Next offers a broad set of generation and processing workflows alongside multiple hardware options. Its security policy describes limited external connections for installation, upgrades, extensions, and requested model downloads.
What the maker publishes
- API access
- Yes source
- Captioning and tagging
- The project lists 25+ built-in LLM and VLM models, OpenCLiP models, and WaifuDiffusion and DeepDanbooru taggers. source
- Captioning and tagging
- The project describes captioning and enhancement with 25+ built-in LLM and VLM models, plus OpenCLIP, WaifuDiffusion, and DeepDanbooru support. source
- Extensions
- SD.Next supports custom scripts and extensions, and its security policy says extensions can be downloaded when enabled. source
- Extensions
- The project supports custom scripts and extensions. source
- External connections
- The security policy says connections may be made for indexes, installation and upgrade packages, enabled extensions, user-requested model downloads from CivitAI or Hugging Face, and user-submitted benchmark information. source
- Guidance tools
- Supported guidance features include LoRA, ControlNet, IPAdapters, and Prompt Enhance. source
- Guidance tools
- The project lists LoRA, ControlNet, IPAdapters, and Prompt Enhance among its guidance features. source
- Hardware support
- The project lists CUDA, ROCm, ZLUDA, Intel IPEX XPU, OpenVINO, DirectML, Apple M1/M2 MPS, ONNX/Olive, and CPU-only execution support. source
- Image editing
- Yes source
- Image processing
- Its processing tools include captioning, tagging, upscaling, interpolation, colorization, filtering, and image color correction. source
- Image processing
- Image processing tools include captioning, tagging, upscaling, interpolation, colorization, filtering, and color grading. source
- Installation
- The project provides installation instructions and quick-start scripts for Linux and Mac, Windows batch files, and PowerShell. source
- Intended users
- The project describes SD.Next as suitable for artists, researchers, and AI enthusiasts. source
- Intended users
- The README describes the application as suitable for artists, researchers, and AI enthusiasts. source
- Interface
- The project describes desktop and mobile interfaces and localization in about 15 languages. source
- License
- The GitHub repository identifies Apache-2.0 as the project license. source
- Memory features
- SDNQ quantization is described as reducing VRAM use by up to 4x, and Balanced Offload balances CPU and GPU memory for larger models on limited hardware. source
- Memory optimization
- Its SDNQ quantization engine can reduce VRAM use by up to 4x, and Balanced Offload balances CPU and GPU memory. source
- Models
- The project says it supports dozens of advanced models and can automatically download selected reference models. source
- Model support
- SD.Next supports a broad range of diffusion models and can automatically download selected reference models. source
- Output formats
- PNG,JPG source
- Platforms
- The README lists Windows, Linux, and macOS support, with GPU acceleration options and CPU-only execution. source
- Privacy
- The security policy says the app does not send or share data externally, track user or system data, store user inputs or detection results, or use analytics services. source
Checked against the maker's pages on 12 Oct 2026.