Audio Transcriber

Overview

Turn recordings into text with OpenAI's Whisper, running right in your browser. Drop in MP3, WAV, M4A, OGG, FLAC, WEBM or MP4 files and get a transcript, with word-level timestamps if you want them. Queue several files and they're done one after another. The model downloads once and stays cached, and your audio never leaves your computer.

Interviews and researchTranscribe interviews or field recordings without handing sensitive material to a cloud service.
Podcast creatorsGet a transcript of each episode for show notes, captions or blog posts.
Lectures and notesTurn lectures or dictated notes into text you can search and edit.
Other languagesLet Whisper detect the language, or pick one yourself.
Meeting notesTranscribe recorded calls and meetings without sharing confidential audio with an outside API.

Key features

Whisper speech-to-text on your device
Transcribe several files in one go
Model download with live progress
Progress for each file
Download each transcript as TXT
Auto-detect the language, or pick one
Optional word-level timestamps
One file at a time to save memory
Audio never leaves your browser

Set

AI Tools 04

Neighbours

01 Run AI in the browser 02 Generate AI Images 03 Generate AI Music 04 Invoice Builder 05 Markdown Editor & Live Preview 06 Spreadsheet Converter

Specification

Command: ai.audio_transcription
Set: AI
Input: text
Output: —
Max file: no limit
Batch: single file

Runtime

Engine: WebLLM · ONNX Runtime Web · WebGPU
Threads: — workers
Isolated: —
WebGPU: —
Memory: —

Transport 0 B

Network use: downloads the model once, never your input.

Keys

Search: ⌘K
Panels: ⌘1 · ⌘2
Full width: F
All keys: ⌘/

How it works

1. Add audio

Drop one or more files, or pick them. MP3, WAV, M4A, OGG, FLAC, WEBM and MP4 all work.

2. Pick a model

Tiny is fastest, Small is most accurate. The model downloads once and stays in your browser.

3. Transcribe

Press Transcribe Audio and Whisper does the rest on your own computer.

Frequently asked

Is my audio sent to a server?

No. Transcription runs in your browser with WebAssembly, so your audio stays on your device.

Does the model download every time?

No. After the first download the model is cached in your browser and reused on later visits.

How large are the models?

Tiny is about 44 MB, Base about 78 MB and Small about 250 MB. Bigger models are more accurate, but slower and need more memory.

Can I transcribe multiple files?

Yes. Add as many as you like and they're transcribed one after another.

What audio formats are supported?

MP3, WAV, M4A, OGG, FLAC, WEBM and MP4. Each file is converted to the 16 kHz audio Whisper needs.

What browsers are supported?

Chrome, Edge, Firefox and Safari all work. Chrome and Edge are fastest, because they can use WebGPU.

Can I transcribe in languages other than English?

Yes. Pick a language, or leave it on Auto-detect and Whisper works it out.

Keys

Search: ⌘K
Panels: ⌘1 · ⌘2
Full width: F
All keys: ⌘/