BREAKING
TinyTTS: 1.6M-param speech model
0
M
parameters
0
MB
ONNX FP16
0
x
real-time CPU
CPU real-time factor by model
TinyTTS ONNX
53
Piper
26
Kokoro
3
Parameter count vs rivals
TinyTTS
1.62
Piper
63
Kokoro
82
Strengths and current limits
Strengths
●
Apache 2.0, VITS-based
●
44.1 kHz audio output
●
Python and Node.js installs
Limits
●
English-only
●
No zero-shot voice cloning
●
Training code not released
Tiny footprint, big speed
AI NEWS BLITZ
A new open-source text-to-speech model runs entirely on CPU with just 1.6 million parameters.