BREAKING
Soprano: 80M On-Device TTS Model
0
M
parameters
0
x
real-time CPU
0
GB
under memory
0
%
fewer hallucinations
0
%
user preference
0
+
HF downloads
Runs locally, ships with tools
1
pip install soprano-tts
↓
2
WebUI, CLI, API
↓
3
ONNX & ComfyUI
↓
4
CPU, CUDA, MPS
Strengths and current limits
Strengths
●
Sub-15ms GPU streaming
●
Runs on a Raspberry Pi
●
Apache-2.0 license
Limits
●
English-only, no cloning yet
●
~1,000 hours of training data
●
No large benchmark paper
Edge TTS gains momentum
AI NEWS BLITZ
An open-source text-to-speech model called Soprano is drawing developer attention for running fully on-device.