AI Runner
An open-source desktop AI app for Linux that runs chat, voice and image models locally, with a layered canvas and a self-hosted API.
AI Runner combines a personal AI companion with a layered art canvas, running both on your own hardware. It's for people who want private voice conversations and image creation in the same desktop app. The local experience works offline without an API key or subscription, and conversations, memories and companion settings stay on your machine.
You can give the companion a name, personality and voice. It remembers past conversations across sessions and uses that memory in later replies. Speech recognition and text-to-speech let you talk aloud, while its mood changes through conversation.
The canvas supports drawing, painting and compositing on layers. You can turn sketches into images, revise existing images or use inpainting to change selected areas. Supported image models include Stable Diffusion 1.5, SDXL and Z-Image Turbo, with LoRA support, filters and background removal.
For local LLM chat, AI Runner supports GGUF models through llama.cpp, including Qwen3.5-9B, and can connect to Ollama. OpenAI and OpenRouter are optional cloud providers. It has no telemetry; data leaves your machine only when you explicitly enable a cloud feature. Built-in downloaders fetch models from HuggingFace and Civitai.
Linux is the primary supported platform. The stated minimum hardware includes 16 GB RAM and an NVIDIA RTX 3060. AI Runner is free and open source under GPL-3.0.
You can also run it as a self-hosted server for chat, image generation and speech. Its HTTP API includes Ollama-compatible endpoints and OpenAI-compatible chat completions with tool support for application integrations and automated workflows.
