DeepInfra icon

DeepInfra

Use DeepInfra API for AI operations

Actions6

Speech Recognition → Transcribe

AI-generated

Overview

This node integrates with the DeepInfra API to perform speech recognition tasks, specifically transcribing or translating audio content. It supports input either as a direct URL to an audio file or as binary audio data provided within the workflow. The node sends the audio to a selected OpenAI Whisper model variant for transcription or translation.

Common scenarios include:

  • Automatically converting recorded meetings, interviews, or podcasts into text.
  • Translating spoken content from one language to another.
  • Processing audio files stored remotely via URLs or uploaded directly in workflows.

Practical example:

  • A user provides a URL to a podcast episode and selects the "Transcribe" operation with the "Openai Whisper-Large-V3" model to generate a text transcript of the episode.
  • Alternatively, a user uploads an audio recording as binary data and uses the "Translate" operation to get the translated text output.

Properties

Name Meaning
Model The speech recognition model to use. Options: "Openai Whisper-Large-V3", "Openai Whisper-Large-V3-Turbo".
Input Type Specifies whether the audio input is provided as a URL or as binary data. Options: "URL", "Binary Data".
Audio URL (Required if Input Type is URL) The URL pointing to the audio file to be transcribed or translated.
Binary Property (Required if Input Type is Binary Data) The name of the binary property containing the audio data.
Options Additional optional parameters:
- Language: ISO-639-1 code specifying the audio language.
- Prompt: Optional guiding text to influence transcription style or continuation.
- Temperature: Sampling temperature controlling randomness in transcription (0 to 1).

Output

The node outputs a JSON object containing the transcription or translation result returned by the DeepInfra API. The exact structure depends on the API response but generally includes the recognized text and metadata about the transcription.

No binary output is produced by this node.

Example output snippet (simplified):

{
  "text": "Recognized speech text here",
  "language": "en",
  "duration": 123.45,
  ...
}

Dependencies

  • Requires an active DeepInfra API key credential configured in n8n.
  • Uses the DeepInfra OpenAI-compatible API endpoint at https://api.deepinfra.com/v1/openai.
  • Downloads audio files locally when input is a URL or reads binary data from workflow input.
  • Node depends on standard Node.js modules (fs, path, os) and external libraries (axios for HTTP requests, openai SDK).

Troubleshooting

  • Audio file access issues: If using a URL input, ensure the URL is publicly accessible and returns valid audio data. Private or protected URLs may cause download failures.
  • Invalid binary data: When using binary input, verify that the specified binary property contains valid base64-encoded audio data.
  • Model selection errors: Confirm the chosen model is supported and correctly spelled.
  • Language code problems: Use valid ISO-639-1 codes for the language option; otherwise, transcription quality may degrade or fail.
  • Temporary file cleanup: The node writes audio to temporary files during processing. In rare cases, leftover temp files might accumulate if the node crashes before cleanup.
  • API errors: Network issues or invalid API keys will cause errors. Check credentials and network connectivity.
  • Error messages: Errors thrown typically contain descriptive messages from the DeepInfra API or internal node checks. Enabling "Continue On Fail" allows partial processing of multiple items.

Links and References

Discussion