Groq Speech-to-TextInstall

Summary

Transcribe audio files to text by submitting either a URL or binary audio data to the Groq Whisper API, selecting model, language, and response format options.

Inputs

• Input Type (required) — Specify whether the audio file is provided as a URL or as binary data in the workflow.
• Audio URL (required) — If Input Type is 'Audio URL', provide the URL pointing to the audio file to transcribe (supported formats: mp3, mp4, wav, m4a, flac, ogg, webm).
• Binary Property (required) — If Input Type is 'Binary Data', specify the name of the binary property containing the audio file within the incoming items.
• Model (required) — Choose the Whisper model to use for transcription; options include 'Whisper Large V3 Turbo' for faster, lower-cost transcription, or 'Whisper Large V3' for higher accuracy.
• Language — Optionally set the audio language as an ISO-639-1 code to improve transcription accuracy; if omitted, language auto-detection is applied.
• Response Format (required) — Select the format of the transcription response: simple JSON (text only), verbose JSON (with metadata and timestamps), or plain text.
• Additional Options — Optionally provide a prompt to guide transcription, set the sampling temperature (0-1), and include timestamp granularities at segment and/or word level.

Output shape

a list of JSON objects, one per input item, each containing the transcription result in the requested format along with metadata fields

The output JSON includes the transcription data under either 'text' (if plain text requested) or full response structure, plus metadata with model used, operation type, detected or specified language, input type, and filename. Errors return an object with an 'error' message.