Overview

Convert speech to text with Whisper. The workflow returns plain text, with no timestamps.

  1. Audio Input (nodetool.input.AudioInput) - Records or uploads audio
  2. Automatic Speech Recognition (nodetool.text.AutomaticSpeechRecognition) - Transcribes the audio. The template selects openai/whisper-large-v3 on fal.ai. The node also takes a language code (empty detects it), a guiding prompt, and a temperature.
  3. Output - The transcript text

For timed segments or words, use openai.audio.Transcribe, which returns text, segments, and words.

Demo

How to Use

Record or upload audio on the input node, then run. The output node holds the transcript.

Tags

start, audio, asr

Workflow Diagram

graph TD audio["AudioInput (audio)"] asr["AutomaticSpeechRecognition"] transcript["Output (transcript)"] audio --> asr asr -->|text| transcript