AI utilities
Speech to Text
Transcribe microphone or audio files to text in your browser with Whisper ONNX — private, free, no upload.
- Mobile Friendly
- Accessibility Tested
- SEO Optimized
- Responsive
Version 1.0.0 · Last reviewed
Speech to Text
On-device Whisper Tiny (ONNX). Prefer WebGPU, fall back to WebAssembly. Audio stays in your browser.
Drag & drop audio, or choose a file
MP3, WAV, M4A, WebM, OGG, FLAC
Microphone
Ready to record
No transcript yet
Start recording or upload an audio file to see live and final text here.
How it worksShow guideHide guide
Introduction
Speech to Text turns microphone recordings or audio files into editable text entirely in your browser. It uses a Whisper ONNX model via Transformers.js and onnxruntime-web — prefer WebGPU when available, otherwise WebAssembly — with no uploads and no cloud AI APIs.
Step-by-step
- Open this page — the Whisper Tiny model loads only here (not on the homepage).
- Allow microphone access to record, or drop / upload an audio file (MP3, WAV, M4A, WebM, OGG, FLAC).
- Watch download and inference progress; the model is cached for later offline reuse.
- Edit the transcript, then copy or download TXT / Markdown.
Worked example
Record a 30-second voice memo, wait for the first-run model download, then review the live transcript. Pause if needed, resume, and stop when finished. Copy the text or download a Markdown file with optional timestamps.
Use cases
- Dictate notes, emails, and drafts hands-free
- Transcribe interviews, lectures, and meeting recordings
- Generate starting text for captions or show notes
- Convert voice memos into searchable documents offline (after the first model download)
Frequently asked questions
Is my audio uploaded?
No. Recognition runs on-device with Transformers.js and onnxruntime-web. Audio never leaves your browser for transcription.
Why does the first visit download a model?
Whisper Tiny ONNX weights download once and are stored in the browser cache. Later visits reuse the cached model unless a new model version is deployed.
WebGPU or WebAssembly?
The tool prefers WebGPU for speed when your browser supports it, and automatically falls back to WebAssembly (CPU) otherwise.
Which languages are supported?
The multilingual Whisper Tiny model can auto-detect language for many locales. Accuracy is best for clear speech and common languages.
Can I use this offline?
After the model has downloaded once, transcription can run without a network connection. The page itself still needs to be loaded from the site or a cached visit.
Related tools
Word Counter
Count words, characters, sentences, paragraphs, and estimated reading time — privately in your browser.
Markdown ↔ HTML
Convert Markdown to HTML and HTML back to Markdown — locally in your browser.
Case Converter
Convert text to UPPERCASE, lowercase, Title Case, sentence case, camelCase, snake_case, or kebab-case.
Popular tools
Average Calculator
Calculate the mean, sum, median, and range from a list of numbers.
BMI Calculator
Calculate Body Mass Index (BMI) from height and weight, and see the WHO adult category.
BMR Calculator
Estimate basal metabolic rate with Mifflin–St Jeor and see rough daily calorie needs by activity level.
Was this helpful?
Your feedback stays on this device only — it is not sent to the Bestrao team yet.