Text-to-speech

wfloat/wfloat-tts

Model ID: wfloat/wfloat-tts
Type: Text-to-speech

Use this model

Use in a web app

Install the browser SDK:

npm install @wfloat/wfloat-web

Load the model and synthesize an utterance. The first load downloads the model and browser runtime; later loads use the browser cache.

import { loadTtsModel } from "@wfloat/wfloat-web";

const tts = await loadTtsModel("wfloat/wfloat-tts", {
  onProgress(event) {
    console.log(event.status, event.progress ?? "");
  },
});

const result = await tts.synthesize({
  text: "The signal is clean. Start the recording.",
  voice: "narrator_woman",
  emotion: "neutral",
  intensity: 0.5,
  speed: 1,
  silencePaddingSec: 0.1,
});

console.log(result.audio.sampleRate);
console.log(result.timeline.chunks);

Inference and playback run on the user’s device. Use tts.pause(), tts.play(), and tts.stop() to control the active request.

Use in a React Native app

Install the React Native SDK:

npm install @wfloat/react-native-wfloat

Run pod install from the app’s ios directory after installing. Android is connected through React Native autolinking.

import { loadTtsModel } from '@wfloat/react-native-wfloat';

const tts = await loadTtsModel('wfloat/wfloat-tts', {
  onProgress(event) {
    console.log(event.status, event.progress ?? '');
  },
});

const result = await tts.synthesize({
  text: 'All systems are stable. Begin the launch sequence.',
  voice: 'narrator_woman',
  emotion: 'neutral',
  intensity: 0.5,
  speed: 1,
  silencePaddingSec: 0.1,
});

console.log(result.audio.sampleRate, result.audio.durationSec);
console.log(result.timeline.chunks);

The model runs locally on iOS and Android. Use tts.pause(), tts.play(), and tts.stop() to control the active request.

Use in Python

Install the Python SDK:

pip install wfloat

Load the model, synthesize speech, and save the result as a WAV file:

import wfloat

tts = wfloat.load_tts_model("wfloat/wfloat-tts")

result = tts.synthesize(
    text="No, that is not possible.",
    voice="mad_scientist_woman",
    emotion="surprise",
    intensity=0.7,
    speed=1.0,
    silence_padding_sec=0.1,
)

result.audio.save("speech.wav")
print(result.model_id)
print(result.timeline.chunks[0].text)

The first load downloads the required assets. Subsequent loads use the local model cache.