Woolverine94/biniou

a self-hosted webui for 30+ generative ai

What it solves

Biniou is a self-hosted web interface that allows users to run various generative AI models locally on their own hardware. It eliminates the complexity of setting up multiple disparate AI tools by providing a unified dashboard for text, image, audio, video, and 3D generation, and it is designed to work even on machines without a dedicated GPU (starting from 8GB RAM).

How it works

Powered by Hugging Face and Gradio, Biniou integrates a wide array of AI modules. It uses llama-cpp for text-based chatbots and supports a vast range of image generation models (including Stable Diffusion, Flux, and Kandinsky) and audio/video tools. The system handles the automatic installation of prerequisites and models, allowing users to deploy it across Linux, Windows, and macOS. It also supports hardware acceleration via CUDA and ROCm for those with compatible GPUs.

Who it’s for

Users who want a private, offline-capable GenAI suite on their own computer without needing high-end professional hardware or deep technical knowledge of model deployment.

Highlights

  • Multimodal Capabilities: Supports text (chatbots, translation, speech-to-text), images (generation, inpainting, outpainting, faceswapping), audio (music and speech generation), video (text-to-video), and 3D objects.
  • Low Hardware Barrier: Can run on systems with as little as 8GB of RAM and no dedicated GPU.
  • Unified Interface: A single web UI with a control panel for updating, restarting, and managing models.
  • Broad Model Support: Compatible with GGUF quantized models for LLMs and various formats (safetensors) for image generation, including LoRA and textual inversion.
  • Cross-Platform: One-click installers available for multiple Linux distributions, Windows, and experimental support for macOS.