Skip to main content

transcribe()v4.0.518

Transcribes a 16kHz mono waveform and returns words with start and end timestamps.

transcribe.ts
import {resampleTo16Khz, transcribe} from '@remotion/whisper-webgpu';

const file = new File([], 'audio.wav');
const channelWaveform = await resampleTo16Khz({file});

const result = await transcribe({
  channelWaveform,
  model: 'small.en',
});

console.log(result.text, result.words);

Options​

channelWaveform​

A mono Float32Array sampled at 16kHz. Use resampleTo16Khz() for browser audio files.

model​

One of the timestamped models returned by getAvailableModels(). small or small.en is the recommended default.

language?​

The spoken language name or language code. This option is required for multilingual models because automatic language detection is not supported. Omit it for English-only models such as small.en.

task?v4.0.523​

Either transcribe or translate. Default: transcribe.

translate translates speech into English and is supported by multilingual non-turbo models only. Check the model's supportsTranslation property before using it. Word timestamps may be less reliable when translating.

chunkLengthInSeconds?​

Length of long-audio chunks. Default: 30. Must be a finite number greater than 0.

strideLengthInSeconds?​

Overlap on both sides of a chunk. Default: 5. Must be finite, non-negative, and less than half of chunkLengthInSeconds.

forceFullSequences?v4.0.523​

Whether every sequence must end with a timestamp. Default: false. Must be a boolean. When set to true, transcription throws if the model returns a trailing sequence without an ending timestamp.

doSample?v4.0.523​

Whether to sample tokens probabilistically instead of always choosing the most likely token. Default: false. Must be a boolean. Sampling can produce different results between runs.

temperature?v4.0.523​

Controls randomness when doSample is true. Default: 1. Must be a finite number greater than 0; lower values favor more likely tokens.

topK?v4.0.523​

Limits sampling to the tokens with the highest probabilities when doSample is true. Default: 50. Must be a non-negative integer; 0 disables the limit.

repetitionPenalty?v4.0.523​

Adjusts the probability of tokens that have already been generated. Default: 1, which applies no penalty. Must be a finite number greater than 0; values above 1 discourage repetition and values below 1 encourage it.

noRepeatNgramSize?v4.0.523​

Prevents a generated token sequence of this length from occurring more than once. Default: 0, which disables the constraint. Must be a non-negative integer.

onModelLoadProgress?​

Called while the model is loaded. To load independently, use loadWhisperModel().

Return value​

Returns the full text, timestamped words, and selected model.

Word-level timestamps are returned after transcription completes. Transformers.js does not currently stream word-aligned updates while inference is running.

Compatibility​

BrowsersEnvironments
Chrome
Firefox
Safari

See also​