Free Audio Transcribe

Transcribe audio files using free local transcription

Overview

This node performs audio transcription using a free local transcription engine based on the Whisper model. It processes audio files provided as binary data and converts their spoken content into text. This is useful for automating transcription tasks such as converting meeting recordings, interviews, podcasts, or any audio content into searchable and editable text.

Practical examples:

  • Transcribing customer support calls to analyze conversations.
  • Converting lecture recordings into text notes.
  • Generating subtitles or captions from video audio tracks.

Properties

Name Meaning
Binary Property Name of the binary property that contains the audio file to transcribe (e.g., "data").
Language Language of the audio file. Options: Auto-detect, English, German, Spanish, French, Italian, Japanese, Chinese.

Output

The node outputs an array of items where each item contains:

  • json.transcription: The transcribed text result of the audio file.
  • All original JSON data from the input item is preserved alongside the transcription.
  • The original binary data (audio file) is also preserved in the output.

No binary output is generated by this node; it only produces textual transcription results.

Dependencies

  • Requires the external Node.js package whisper-node which is installed automatically if not present.
  • Uses a local Whisper model file named whisper-tiny.bin stored under a models directory relative to the node's source code.
  • No external API keys or online services are required since transcription runs locally.

Troubleshooting

  • Missing binary data: If the specified binary property does not exist or lacks a file path, the node will throw an error indicating the missing audio file. Ensure the binary property name matches the input data and that the audio file is accessible.
  • Installation failure: If automatic installation of the whisper-node package fails, the node will report an error. Manually installing the package in the n8n environment may resolve this.
  • Model file issues: The node expects the Whisper model file to be present or creates the models directory if missing. If transcription fails due to model loading, verify file permissions and disk space.
  • Unsupported audio formats: The node relies on the underlying Whisper implementation; unsupported or corrupted audio files may cause errors.

Links and References

Discussion