Auto Subtitles for Video
Most people watch social videos with the sound off. Add captions automatically: the speech is transcribed by AI, you can correct any line, and then either burn the captions into the video or download a subtitle file. The AI runs in your browser, so your video is never uploaded.
How to add subtitles to a video
- 1Click “Choose video” or drag a video onto the page, and pick the spoken language (or let it detect it).
- 2Click “Generate subtitles” and fix any words in the transcript.
- 3Pick a caption style and click “Add captions to video”, or download an SRT or VTT file.
Burned-in captions or subtitle files
Burned-in captions become part of the picture, so they show everywhere, including Instagram, TikTok and WhatsApp. SRT and VTT files are separate subtitle tracks for YouTube, video editors and players that let viewers switch them on and off.
Caption styles
“Classic box” puts white text on a dark box for easy reading. “Bold” uses large yellow capitals with an outline, the look popular on Reels and Shorts. “Minimal” is plain white text with a thin outline. Captions can sit at the bottom or in the middle of the frame.
Private by design
Speech recognition uses OpenAI's Whisper model running on your own device. The model downloads once (about 80 MB for Fast, 250 MB for More accurate) and is then cached; your video is never uploaded.
Frequently asked questions
- Is it free?
- Yes, with no signup and no watermark.
- Is my video uploaded?
- No. Transcription and captioning happen in your browser.
- Which languages work?
- Around two dozen common languages can be chosen directly, including English, Hindi, Spanish, Portuguese, French, German and Japanese, or the language can be detected automatically.
- What if a word is wrong?
- Click any line in the transcript and type the correction before adding the captions or downloading the file.
- How long does it take?
- After the one-time model download, transcription usually takes less time than the video lasts on a modern computer. Burning captions re-encodes the video, which takes roughly as long again.