How it works
- Drop the video. The sound is extracted on your computer.
- Create subtitles. The first time, the speech model is downloaded (about 80 MB for Fast, 250 MB for Accurate) and cached by your browser — next time it starts immediately. Recognition then runs on your processor; your audio never leaves the device.
- Edit. Click a time to jump there, correct the text in the list, delete lines, or shift all subtitles if they are early or late.
- Save. Download an .srt or .vtt file, or save the video with the subtitles burned into the picture.
Burned-in or a separate file?
- Burned in (open captions) — part of the picture, visible everywhere: Instagram, TikTok, LinkedIn, WhatsApp, any TV. Most social videos are watched with the sound off, so this is the usual choice for social media. The video is re-encoded.
- .srt next to the video (closed captions) — viewers can turn them on or off. Name it like the video (
holiday.mp4+holiday.srt) and VLC, Plex, Kodi and most smart TVs pick it up automatically. YouTube and Vimeo accept SRT and VTT uploads too.
How accurate is it?
For clear speech in a quiet room, Whisper gets most words right; names, brands and technical terms are where you will make corrections. Background music, several people talking at once and strong accents lower the accuracy. “Accurate” uses a larger model and makes noticeably fewer mistakes, at about a third of the speed.
How long does it take?
On a typical laptop from the last few years, the Fast model processes speech about 3–4 times faster than real time — a 10-minute video in roughly 3 minutes. Accurate runs at about real time. Older or low-power computers are slower; you can keep using other tabs while it works.
Languages and translation
Whisper recognises about 100 languages and detects the language by itself. It can also translate speech into English subtitles: choose “English (translate)” — useful for sharing a video recorded in another language with an English-speaking audience.
Private by design
Most subtitle generators upload your video to their servers and charge per minute. Here the model runs locally in your browser, so there is no upload, no time limit and no subscription — useful for interviews, internal meetings and anything confidential.
Questions
Is it free?
Yes, with no minute limits, no watermark and no sign-up.
Is my video uploaded?
No. Only the speech model is downloaded once from Hugging Face; your video and audio stay on your computer.
What is the difference between SRT and VTT?
Both are plain-text subtitle files. SRT is supported by almost every player and TV; VTT is the web standard used by HTML5 players. Download whichever your platform asks for.
Can I add my own SRT file and burn it in?
Yes — use “Import SRT/VTT”, check the timing, and save the video with burned-in subtitles.
Why is the first run slow?
The speech model has to be downloaded once (80–250 MB). After that it is cached by your browser.
Last updated: 2026-10-30