license: apache-2.0
language:
- en
pipeline_tag: text-to-speech
LuxTTS (mirror)
This is an unmodified mirror of YatharthS/LuxTTS.
All credit for the model belongs to its author, YatharthS
— original code and usage docs: https://github.com/ysharma3501/LuxTTS.
We mirror it only to pin a dependency: gloam.fm downloads this model during first-run setup, so an upstream rename or deletion would break setup for every new user. The weights here are byte-identical to upstream. If you are looking for LuxTTS itself, use the original repo — it is the canonical source and the place to star, file issues, and follow updates.
Released under Apache-2.0, same as upstream.
The upstream model card follows, unmodified.
LuxTTS
This is the model for LuxTTS, a lightweight zipvoice based text-to-speech model designed for high quality voice cloning and realistic generation at speeds exceeding 150x realtime.
Main features
- Voice cloning: SOTA voice cloning on par with models 10x larger.
- Clarity: Clear 48khz speech generation unlike most TTS models which are limited to 24khz.
- Speed: Reaches speeds of 150x realtime on a single GPU and faster then realtime on CPU's as well.
- Efficiency: Fits within 1gb vram meaning it can fit in any local gpu.
Details
- Based on ZipVoice, distilled to 4steps.
- Uses 48khz vocoder instead of 24khz vocoder.
- Implemented higher quality sampling technique then standard euler.
Usage
Please check out the repo for usage: https://github.com/ysharma3501/LuxTTS.git
License
Model and code is released under Apache-2.0 license.