Speech-to-text

k2-fsa/streaming-zipformer-en

Model ID: k2-fsa/streaming-zipformer-en
Type: Speech-to-text

Use this model

Stream microphone transcription in a web app

Install the browser SDK:

npm install @wfloat/wfloat-web

Load the streaming model, create a session, and let the SDK capture microphone audio:

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

const stt = await loadSttModel("k2-fsa/streaming-zipformer-en");
const session = await stt.createSession();

await session.startMicrophone({
  sampleRate: 16000,
  onResult(partial) {
    console.log(partial.text, partial.isEndpoint);
  },
});

// Call these when the user stops recording.
await session.stopMicrophone();
const result = await session.finish();
console.log(result.text);

await session.close();

For an application-owned audio pipeline, use session.push({ audio, sampleRate }) and read partial results with session.getResult().

Stream microphone transcription in React Native

Install the React Native SDK:

npm install @wfloat/react-native-wfloat

Run pod install from the app’s ios directory, and configure microphone permission for the target platform. Then open a streaming session:

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

const stt = await loadSttModel('k2-fsa/streaming-zipformer-en');
const session = await stt.createSession();

await session.startMicrophone({
  sampleRate: 16000,
  onResult(partial) {
    console.log(partial.text, partial.isEndpoint);
  },
});

// Call these when the user stops recording.
await session.stopMicrophone();
const result = await session.finish();
console.log(result.text);

await session.close();

For application-owned PCM audio, feed chunks with session.push({ audio, sampleRate }).

Stream transcription in Python

Install the Python SDK:

pip install wfloat

Load the model and create a streaming session:

import wfloat

stt = wfloat.load_stt_model("k2-fsa/streaming-zipformer-en")
session = stt.create_session()

try:
    # pcm_chunk is a sequence of mono floating-point samples.
    session.push(pcm_chunk, sample_rate=16000)

    partial = session.get_result()
    print(partial.text)

    final_result = session.finish()
    print(final_result.text)
finally:
    session.close()

Call session.push(...) repeatedly as audio arrives, rather than waiting for the complete recording.