WaTale: A free, fully local visual novel engine (Powered by SD 1.5, LayerDiffuse, and ControlNet)
Generate branching visual novels privately with real-time AI images and voice.
Developer Churrucaman has launched WaTale, a free, fully local visual novel engine that combines text, image, and voice AI to create interactive branching narratives entirely on your own hardware. The engine leverages Ollama for text generation, Stable Diffusion 1.5 for images using LayerDiffuse and ControlNet, and Kokoro ONNX for text-to-speech, ensuring all data remains private. There is also optional support for cloud text models via Ollama Cloud, Anthropic, or OpenAI APIs.
WaTale handles real-time generation and playback, rendering SD-generated scene backgrounds with depth parallax, full-body transparent character sprites with idle animations, and real-time lip-syncing through face inpainting. Users can create custom characters, insert themselves into stories, play through generated narratives with integrated minigames, export stories, or let characters interact autonomously. This is an early preview requiring an NVIDIA GPU with at least 4GB VRAM, so some bugs may occur. Feedback is sought, especially on the Stable Diffusion implementation.
- Uses Ollama for text, SD 1.5 with LayerDiffuse/ControlNet for images, and Kokoro ONNX for TTS
- Real-time rendering with depth parallax backgrounds, transparent sprites, and lip-syncing via face inpainting
- Free and fully local, requiring NVIDIA GPU with 4GB+ VRAM; optional cloud API support
Why It Matters
WaTale democratizes interactive storytelling by running entirely locally, ensuring privacy and creative control for users.