biniou

biniou brings AI chat and media creation into one browser interface on your own computer or server. It's for people who want to work with several kinds of generative AI without switching between separate apps.

• Chat and language: Run GGUF models through llama-cpp-python, discuss images with LLaVA, transcribe speech with Whisper, and translate text with NLLB.
• Image creation and editing: Generate images with Stable Diffusion, Flux, Kandinsky or PixArt, then use inpainting, outpainting, ControlNet, upscaling and face restoration to edit them.
• Music and speech: Create music with MusicGen, generate sound effects with AudioGen, or produce spoken audio with Bark.
• Video and 3D: Generate or edit video with AnimateDiff, Stable Video Diffusion and Video Instruct-Pix2Pix, and create 3D objects from text or images with Shap-E.
• Connected workflows: Pass one module's output into another and retain generation settings in the resulting files' metadata.

The open-source app uses the GPL-3.0 license; individual models have separate licenses. It runs on Linux, Windows and Docker, with experimental, untested native installation on Intel Macs. It works offline after installation and model downloads. Models and generated files stay on the host.

An AMD64 CPU and 8GB RAM are the minimum, though some modules require at least 16GB. CPU-only operation is supported but can be slow; NVIDIA CUDA acceleration is available. The default models across all modules need roughly 200GB of disk space. Apple Silicon isn't supported natively.