Capsize-Games/airunner

What it solves

AI Runner provides a private, offline environment for interacting with AI companions and generating AI art. It eliminates the need for internet connections, API keys, or subscriptions by running everything locally on the user's hardware, ensuring that conversations, memories, and creations remain private.

How it works

The project combines a customizable AI companion with a layered art canvas. The companion uses local LLMs, Text-to-Speech (TTS), and Speech-to-Text (STT) to engage in voiced conversations with long-term memory powered by RAG. The art canvas integrates image generation models like SDXL and Z-Image Turbo, allowing users to sketch, paint, and iterate using image-to-image and inpainting. It can also run as a headless HTTP API server compatible with Ollama and OpenAI endpoints for remote access or integration with other tools.

Who it’s for

Users who want a highly private, local AI experience for companionship and creative art generation without relying on cloud services.

Highlights

• Privacy-First Design: Fully local execution with no external APIs or telemetry.
• Multimodal Capabilities: Integrated voice conversation (TTS/STT) and image generation.
• RAG-Powered Memory: The companion builds long-term memory of the user across sessions.
• Layered Art Canvas: A drawing surface that supports sketching, compositing, and image-to-image iteration.
• Flexible Deployment: Available as a GUI application or a headless API server with Ollama/OpenAI compatibility.
• Model Management: Built-in downloaders for HuggingFace and CivitAI models.