BREAKING
LuxTTS: Voice Cloning in 1GB VRAM
0
x
realtime on GPU
0
GB
VRAM footprint
0
kHz
audio output
How LuxTTS Was Built
1
Based on ZipVoice
↓
2
Distilled to 4 steps
↓
3
48kHz vocoder
↓
4
Zero-shot cloning
0
s
reference audio
0
MB
min model size
0
MB
max model size
Local Alternative to Cloud TTS
LuxTTS
Apache-2.0
●
Runs in 1GB VRAM
●
CUDA, CPU and MPS
●
Free and open source
Commercial TTS
●
Cloud-dependent
●
Paid access
●
Less local control
Small Model, Big Reach
AI NEWS BLITZ
A new open-source TTS model clones voices locally in just one gigabyte of VRAM.