Drop any video or audio file and get a word-timed transcript you can correct with a click, then export as SRT, VTT, ASS or plain text. No sign-up, no minute caps, no upload: the AI runs on your device.
Generate an SRT freeLive preview, rendered on this page by the same engine that burns your video.
MP4 and MOV video, or MP3, WAV, M4A and other audio. The file opens directly in the page and is never uploaded, which you can verify in the network tab; the only sizeable download is the one-time model fetch.
Whisper times every single word. Click any word to correct it, double-click to jump the playhead there. Names, brands and jargon take seconds to fix before they reach the file.
SRT for YouTube and social uploads, VTT for the web, ASS for typesetting tools, or plain text for show notes. Or burn the captions straight into a new MP4 on the main page.
SRT (SubRip Subtitle) is the plain-text standard the whole industry accepts: numbered cues, start and end timecodes, caption text. YouTube, Instagram, LinkedIn, Vimeo, Premiere Pro, DaVinci Resolve and Final Cut all take it. Platforms index subtitle text, viewers watching on mute keep reading, and accessibility improves. One small file, three jobs done.
| What matters | Burncaps | Typical online SRT site |
|---|---|---|
| Price | Free, unlimited | Free trial minutes, then a subscription |
| Length limits | None | Minutes-per-month caps |
| Sign-up | None | Account before download |
| Your file | Stays on your device | Uploaded to their servers, queued |
| Word-level editing | Click any word, then export | Varies; often behind the paywall |
No. Because transcription runs on your own device, there is no server bill to protect with minute caps. An hour-long podcast episode works the same as a 30-second clip; longer files simply take longer to process on your machine.
SRT, VTT (the web-native subtitle format), ASS (used by advanced players and typesetting tools) and plain text. You can also burn the captions into the video itself as a new MP4.
It uses OpenAI's open-source Whisper model with word-level timestamps. Clear speech in English transcribes very well; noisy audio and other languages benefit from the Pro max-accuracy model. Every word is click-to-edit before export, so the final file is exactly what you approve.
No. The model downloads into your browser once (about 80 MB, cached), then everything runs locally. Your file never leaves your device, which you can verify in the network tab.
About 100, with auto-detect on by default. You can pick the language manually for better results on accented or mixed speech.