Audio & Video to Text — Transcribe and Make Subtitles (SRT)

Drop an MP3, WAV, M4A, MP4, MOV or WebM file and Localvert writes down what is said, with a timestamp for every line. Copy the text, or download it as SRT or VTT subtitles for YouTube, video editors and media players, or as plain text for notes.

The speech recognition model is Whisper, an open-source model trained on hundreds of thousands of hours of speech in many languages. It runs inside your browser through WebAssembly, so your interviews, lectures, meetings and voice notes never leave your device.

Two model sizes are available. Fast downloads about 40 MB and finishes sooner; it is fine for clear speech in English. Accurate downloads about 77 MB and makes noticeably fewer mistakes, especially in other languages, with accents or with background noise. Either model is downloaded once and kept in your browser for next time.

Choosing the spoken language instead of automatic detection usually improves the result, particularly for short recordings or when several languages are mixed.

Transcription runs on your own processor. On a recent laptop a minute of audio typically takes from several seconds to under a minute; phones and older computers are slower. The progress bar moves in 30-second steps. Keep the tab open until it finishes — closing it stops the work, because nothing runs on a server.

Clear audio matters more than anything else: one person speaking close to the microphone transcribes far better than a distant voice in a noisy room. Music and overlapping speakers reduce accuracy. Always read through the transcript before publishing — names, numbers and technical terms are the most common mistakes.