Pocket TTS

Local text to speech for @micdrop/server, running Pocket TTS in your Node process through the sherpa-onnx addon and ONNX Runtime.

Pocket TTS is a 100 million parameter model from Kyutai, designed for the CPU, that takes its voice from a few seconds of reference audio instead of from a catalog. It generates about five times faster than real time and emits its audio while it is still writing the rest, which makes it the local voice with the shortest delay before the first word.

Installation

Install the package:

npm install @micdrop/pocket-tts

Download the weights, converted to ONNX by the sherpa-onnx project:

curl -LO https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/sherpa-onnx-pocket-tts-int8-2026-01-26.tar.bz2
tar xf sherpa-onnx-pocket-tts-int8-2026-01-26.tar.bz2
rm sherpa-onnx-pocket-tts-int8-2026-01-26.tar.bz2

The archive weighs 98 MB and takes 190 MB on disk. A full precision archive, sherpa-onnx-pocket-tts-2026-01-26.tar.bz2, sits next to it and works the same way, since Micdrop looks the weights up by name in the folder and takes the quantized ones first when both files are there.

Usage

import { PocketTTS } from '@micdrop/pocket-tts'
import { MicdropServer } from '@micdrop/server'

const tts = new PocketTTS({
  modelDir: './sherpa-onnx-pocket-tts-int8-2026-01-26',
  voice: 'bria',
})

// Use with MicdropServer
new MicdropServer(socket, {
  tts,
  // ... other options
})

Options

Voices

A few seconds of someone speaking is enough for the model to answer in that voice, with no transcript of the sample and no training step. The archive ships three such recordings in its test_wavs folder, reachable by name through BUNDLED_VOICES, and any wav file works in their place.

import { BUNDLED_VOICES, PocketTTS } from '@micdrop/pocket-tts'

const tts = new PocketTTS({
  modelDir: './sherpa-onnx-pocket-tts-int8-2026-01-26',
  voice: './voices/my-voice.wav',
})

bria is forty seconds of a woman reading calmly. loona is a one second clip, about the shortest a reference can be. frenchAccent is a French speaker, so the recording gives the timbre and the accent while the words come from the text and stay English.

Pocket TTS reads only the first referenceLength seconds, so a long recording costs nothing extra. The model reproduces the quality of the sample along with the voice, which makes a clean recording worth more than a long one.

Languages

Pocket TTS speaks English in this integration. Kyutai also publishes French, German, Spanish, Italian and Portuguese checkpoints, which are all still waiting for an ONNX conversion, so the addon loads the English model. Hand it French and you get fluent nonsense rather than an accent, since it reads every word with English phonemes. Use Piper for the other languages.

Latency

These times come from a MacBook Pro M2, on CPU. The generation timings assume the model is already loaded.

Micdrop cuts the answer into sentences and hands each sentence over in chunks as the model writes it, so the user hears its opening words while the model is still generating its end. The generation stops between two chunks rather than at the end of the sentence, which makes an interruption immediate.

Memory

The model takes about 350 MB once loaded and around 600 MB while generating. That is more than Kokoro at 80 MB, and it stays small next to the language model running beside it. Micdrop loads one instance per configuration and shares it across every call, so a second conversation costs nothing more.

Documentation

Read the guide on running Micdrop with local models to see how Pocket TTS fits with a local LLM and local speech to text.

License

Pocket TTS is released under the MIT license, and the sherpa-onnx addon under Apache 2.0. The reference recordings shipped with the archive carry their own licenses, listed on the Kyutai voice repository.