All tools
Subtitle generator
Make SRT or WebVTT subtitles from a video or recording, on your device. Free, with no account or uploads.
Drop a video or recording to make subtitles
What is said is written down on your device and timed as subtitles, saved as SRT in one press (WebVTT is under More options). Whisper can mishear and make phrases up, so read them through.
Nothing is uploaded · up to 50 files at once
Language: Find it by itself · Listening: Fast
- PrivateWhisper runs in this browser. The sound never leaves your device.
- Works offlineAfter the first time, the model is kept and nothing more is downloaded.
- Word by wordEvery word is timed: the transcript follows the sound, and subtitles read well.
A little help along the way
Follow the steps below. Every option is explained underneath.
- Add your filesDrop them, choose them, paste them or record: MP3, M4A, WAV, FLAC, Opus or videos.
- Choose the language and accuracyWhisper can find the language itself. Fast suits clear speech; Accurate is better with accents and noise, and slower. Translating into English is under More options.
- Watch the words arriveThe transcript window opens and the text comes in as Whisper hears it, with how long is left. Stop at any time and keep what was heard.
- Listen, correct and saveClick any word to hear it, press Correct to fix the text, find and replace, then say what you need (text, subtitles or a Word document) and download it in one press. Every other kind, JSON among them, is under More options.
Questions people ask
- Which subtitle files does it make?
- SRT, chosen for you, which most players and video sites take; WebVTT, for web players, is under More options beside the other kinds of file.
- Are the subtitles always right?
- No. Whisper, the speech model that runs on your device, can mishear words and make phrases up; phrases it may have made up are marked. Read the subtitles through and correct them before you post them.
- Can I correct the subtitles before saving them?
- Yes. Press Correct and type into the transcript: each phrase keeps its times, and the SRT and WebVTT files are made from the corrected words.
- Is my video uploaded to make the subtitles?
- No. The sound is read and heard in this browser. The Whisper model is downloaded from this site once the first time, and nothing about the video is sent anywhere.
- How are the subtitles timed?
- From each word's own time: at most two lines of 42 letters and seven seconds a cue, never running across two phrases.
This page is the Transcribe audio, opened for this one job; every other tool it has is one press away. Each of its settings is explained on the Transcribe audio's own page.
Read the step-by-step guide