Stt

📸 Screenshots

Here are visual examples of this section:

Stt - Worker Configuration Interface

The STT (Speech-to-Text) worker converts audio files containing speech into text transcriptions. It uses OpenAI's Whisper model to provide accurate speech recognition for various audio formats including MP3, WAV, and OGG.

2. Configuration Parameters

The Worker accepts the following parameters:

engine: Specifies the speech recognition engine to use (defaults to "whisper-1")

3. Input/Output Handles

input: Input handle - accepts audio data in multiple formats: base64 encoded audio with extension {audio: string, ext: string}, or URL strings pointing to audio files
output: Output handle - returns the transcribed text as a string

4. Usage Examples with Code

// Configure STT worker with Whisper engine
const sttWorker = {
  parameters: {
    engine: "whisper-1"
  },
  fields: {
    input: { value: { audio: "base64AudioData", ext: "mp3" } },
    output: { value: null }
  }
}

5. Integration Examples

This worker integrates well with Content workers that fetch audio files from URLs, and can feed transcribed text to LLM workers for further processing or analysis.

6. Best Practices

Ensure audio files are in supported formats (MP3, WAV, OGG) for best results
Use clear, high-quality audio recordings to improve transcription accuracy
Consider audio file size limits when processing longer recordings
Verify your OpenAI API key has access to the Whisper API

7. Troubleshooting Tips

If transcription fails, check that the input audio format is supported and properly encoded
Verify the audio file is not corrupted by testing with a different audio player first
For URL inputs, ensure the audio file is publicly accessible and not behind authentication
Check OpenAI API quotas if requests are being rejected

Stt