Whisper Local
Whisper Local turns voice recordings into written text. The transcription runs inside the platform itself: you don’t need an account with an external provider or a credential, and the audio never leaves NappAI.
Use it when you want to turn voice notes, recorded calls or meetings into text so you can summarize, analyze or store them.
When to use it
- Transcribe the voice messages your customers send so an agent can answer them.
- Turn meeting recordings into text and then summarize them with an AI model.
- Process audio that contains sensitive information without sending it to an external service.
- Convert several audio files at once and get a single text with all the transcripts.
Configuration
| Field | What it does |
|---|---|
| Audio (required) | The audio file or files you want to transcribe. Connect it to the component that brings in the audio (an uploaded file, a received message, etc.). It accepts several audios at once. |
| Model size | Size of the transcription model. tiny is the fastest, with basic accuracy. base offers a good balance between speed and accuracy and is the default. |
Advanced options
| Field | What it does |
|---|---|
| Language | Language of the audio. With auto (the default) the component detects it on its own, a little more slowly. If you know the language, pick it to speed things up: en, es, fr, de, it, pt, nl, zh, ja, ko, ar or ru. |
Output
The Transcript output returns the transcribed text as a message. If you send several audios, you get all the transcripts in a single text, separated by a blank line.
You can connect it to an agent, a summarizer, a text formatter or any component that takes text.
Example
A clinic receives voice notes from its patients on WhatsApp. The flow picks up the audio, sends it to Whisper Local with Language set to es, and passes the transcript to an agent that works out whether the patient wants to book, change or cancel an appointment.
Good practices
- If all your audios are in the same language, set it in Language: transcription is faster.
- Start with base. Switch to tiny only if you need more speed and basic accuracy is enough.