metadata
title: README
emoji: 💬
colorFrom: gray
colorTo: yellow
sdk: static
pinned: false
short_description: Model weights for Voxta, speech-first AI companions
Voxta
Voxta is a platform for talking naturally with AI companions, by voice or by text. Speech recognition, text generation, text to speech and image generation are all pluggable, so you can run it entirely on your own hardware or mix in hosted services.
This organization hosts the model weights that Voxta downloads at runtime.
Most of what Voxta uses comes straight from the original model authors, and we point at their repositories directly. We publish here only when a model needs repackaging, quantizing, or a conversion that is not available upstream.
What is here
- ONNX weights for the Ditto talking head model
- Quantized builds of the LLM2Vec text encoder used by the ARDY motion module
- A converted faster-whisper turbo checkpoint used for speech recognition
Links
- Website: https://voxta.ai
- Documentation: https://doc.voxta.ai