Skip to content
Nappai Documentation

OpenAI Whisper

OpenAI Whisper turns audio recordings into text using OpenAI’s transcription service. You can transcribe one or several files at once and use the text in the rest of the flow, for example to summarize it or analyze it with an agent.

It requires an OpenAI account. The audio is sent to OpenAI to be transcribed.

When to use it

  • Transcribe voice notes that arrive by messaging or email.
  • Turn meeting recordings into text so you can summarize them later.
  • Convert customer service calls into text for review.
  • Give an agent the ability to transcribe audio when it needs to.

Configuration

FieldWhat it does
OpenAI API Key (required)Your OpenAI key. You can save it as an OpenAI API credential so you don’t have to type it in every flow. See Credentials.
Audio InputThe audio files you want to transcribe. Connect it to the component that brings in the files, for example a chat with attachments or a storage component. Several files are accepted.
LanguageLanguage of the audio. Choose from en (English, default), es (Spanish), fr, de, it, pt, ja, ko, hi, ar, ru and zh. Set the actual language of the audio for better results.

This component can also work as a tool for an agent. See Tool Mode.

Output

  • Data: one record per file, with the transcript in the text field and the chosen language in language. If a file cannot be transcribed, its record contains an error field with the reason, and the remaining files are still processed.

Example

Your sales team sends voice notes after each visit. You connect the audio files to OpenAI Whisper with Language set to es. The transcript goes to a Summarizer that writes a summary of the visit, and the summary is saved in your customer tool.