Transcribe Audio
Transcribe Audio
Section titled “Transcribe Audio”Turn speech into text entirely in your browser with Whisper (tiny/base, on transformers.js). Private, offline, no cost — audio never leaves the device.
What it does
Section titled “What it does”Runs the audio through an on-device speech-to-text model and returns the transcript plus timestamped segments. The audio is decoded and resampled in the browser before it reaches the model.
When to use it
Section titled “When to use it”Transcribe a voice note, a recorded call, or microphone input captured by Record Audio — then summarize, search, or route the text downstream.
Inputs and settings
Section titled “Inputs and settings”| Setting | Notes |
|---|---|
| Model | Any installed speech-to-text model (e.g. Whisper tiny/base). |
| Audio | Audio URL, data URL, or $binary — pair with Record Audio or a file node. The Record tab lets you record a clip from your microphone right in the settings. |
Record tab
Section titled “Record tab”Next to From input, URL, Upload and Base64, the Audio field has a Record tab: record a clip (up to 2 minutes) from your microphone — you can pick which mic — and it’s stored with the node like an upload. Use it to try a model with Run preview, or as fixed audio for the node. To capture new audio each time the workflow runs, use the Record Audio node instead.
Outputs
Section titled “Outputs”Returns { transcript, segments, chunks }:
transcript— the full text.segments— timed pieces of the transcript as{ start, end, text }, times in seconds. The last segment’sendcan benullwhen the audio cuts off mid-phrase.chunks— the same timing in the model’s raw form ({ timestamp: [start, end], text }), kept for older workflows.
Troubleshooting
Section titled “Troubleshooting”- Empty transcript — check the audio actually contains speech and is a format the browser can decode (e.g.
webm,wav,mp3). - Slow first run — Whisper loads on first use, then stays warm for the session. Prefer
tinyon low-memory devices.