Free, private speech to text
Textpro Speech to Text turns voice notes, meetings, lectures, podcasts and videos into text using OpenAI Whisper, the open-source speech recognition model. Unlike most online transcription services, nothing is uploaded: the model runs on your own device with WebGPU (or WebAssembly on other browsers), so there are no minute limits, no sign-up and no cost.
How it works
- Click Record and speak, or upload or drop an audio or video file.
- Pick a model and language (or leave it on auto-detect). Tick Translate to English to get English text from any language.
- The first time, the model downloads once and is cached. Long files are split into 30-second chunks automatically.
- Copy the text, download a .txt, or download a .srt subtitle file with timestamps.
Which model should I use?
| Model | Download | Speed | Best for |
|---|---|---|---|
| Tiny | ~41 MB | Fastest | Quick English notes, phones |
| Base | ~77 MB | Fast | Everyday use, clearer accents |
| Small | ~249 MB | Slower | Best accuracy, Indian languages, noisy audio |
Sizes are for the CPU build; on WebGPU the GPU build is roughly twice the size but much faster.
Supported languages
| Region | Languages |
|---|---|
| India | Hindi, Tamil, Telugu, Kannada, Malayalam, Marathi, Bengali, Gujarati, Punjabi, Urdu, English |
| Europe | English, Spanish, French, German, Portuguese, Italian, Dutch, Russian, Turkish and more |
| Asia & Middle East | Chinese, Japanese, Korean, Indonesian, Arabic and more |
Whisper supports about 99 languages in total. Accuracy is best for English; for Indian languages, the Small model gives noticeably better results.
Privacy
Your recordings and files are processed only in your browser. The only network requests are for the page itself and the one-time model download from Hugging Face. No audio, text or analytics about your transcripts is sent anywhere.
FAQ
Is this speech to text tool free?
Yes. There are no minutes limits, credits, watermarks or sign-up.
Is my audio uploaded to a server?
No. The Whisper model is downloaded to your browser once and all transcription happens on your device. Your audio and transcripts never leave your computer or phone.
Which languages are supported?
Whisper understands about 99 languages, including English, Hindi, Tamil, Telugu, Kannada, Malayalam, Marathi, Bengali, Spanish, French, German, Arabic, Chinese and Japanese. Pick a language or let it auto-detect.
Can it translate speech to English?
Yes. Tick Translate to English and speech in any supported language is transcribed directly into English text.
Can I make subtitles (SRT) for a video?
Yes. Upload the video file, and after transcription click Download .srt to get a subtitle file with timestamps.
Why is the first run slow?
The first time, your browser downloads the model (about 41 MB for Tiny, 77 MB for Base, 249 MB for Small). It is cached, so later visits start quickly. Browsers with WebGPU (recent Chrome and Edge) transcribe much faster.
Does it work offline?
After the model has been downloaded once, transcription itself runs locally without sending anything over the network.