loadWhisperModel()v4.0.518

Downloads and initializes a timestamped Whisper model. Calling it before transcribe() lets you show model-loading progress.

load-model.ts
import {loadWhisperModel} from '@remotion/whisper-webgpu';

const result = await loadWhisperModel({
  model: 'small.en',
  onProgress: ({progress}) => {
    console.log(progress);
  },
});

console.log(result.alreadyLoaded);

Options

model

One of the names returned by getAvailableModels().

onProgress?

Called while model files are resolved and downloaded. progress is the aggregate download progress between 0 and 1 and never decreases.

Return value

Returns whether the same model pipeline was already loaded in memory.

The browser manages its persistent Transformers.js cache. This function also retains the initialized model in memory until disposeWhisperModel() is called or the page closes.

Compatibility

Browsers			Environments
Chrome	Firefox	Safari	Client-side rendering	Server-side rendering	Player	Studio

See also

• disposeWhisperModel()
• isWhisperModelCached()
• transcribe()