Kokoro
82 million parameters that sound like a thousand times more
What it is
A text-to-speech model that punches absurdly above its weight: natural speech in 9 languages with multiple voices, voice blending, and speed control — small enough to run in real time on a CPU, a phone, or a browser tab.
Why it's interesting
The efficiency outlier of open TTS. At ~82M parameters it remains the default for edge, embedded, and browser deployment, with ONNX, WASM, and Rust ports everywhere you look.
Use cases
- On-device and offline TTS
- Screen readers and accessibility tools
- Low-cost narration at scale
Who it's for
App developers, edge engineers, anyone without a GPU
Setup
Trivial. pip install kokoro (plus espeak-ng); real-time on CPU, under 1GB of memory
Limitations & cautions
No zero-shot cloning — you get a fixed voice set — and less emotional range than the big models. Core development has slowed into maintenance.
Editorial takeaway
Proof that 'small and finished' is a feature. Some models are infrastructure; this one is a brick you build with.