Auto Captions Online

Upload a talking video and get the whole video back with captions burned in — right in your browser. Nothing to install, no API key to set up.

What you get

• captioned.mp4 — your whole video with captions burned in. Nothing is cut.
• captions.srt and captions.vtt — the same captions as files for YouTube, Premiere, CapCut and other editors
• transcript.txt — the plain text, one caption per line

The SRT arrives first, within a couple of minutes, so you can start checking the words while the video renders.

Two formats

• Keep original (default): landscape stays landscape, vertical stays vertical. Nothing is cropped.
• Vertical with black bars: the whole frame in a 9:16 video for Shorts, Reels and TikTok, captions in the lower bar.

Fix captions by chatting

Speech-to-text is not perfect, especially for names, slang and fast Chinese speech. It can mishear words or miss a few. Every fix is one message:

• "Change caption 12 to …"
• "At 1:53 to 1:56 add the sentence …" — for a line that was missed
• "Replace Jon with John everywhere" — for a name heard wrong every time
• "Split caption 5" / "Merge captions 8 and 9" / "Delete caption 20"
• "Show captions 0.3 seconds earlier" — shifts all captions
• "Make the captions bigger" / "Put captions at the top"

After an edit, only the changed part of the video is rendered again, so fixes come back quickly.

Limits

• Chinese or English speech only. Chinese captions are in Traditional Chinese (Taiwan)
• Up to 500 MB, 20 minutes and 1080p per video
• Upload the file itself; links are not supported
• A long video takes a while to render. A 20-minute video may take 15 minutes or more, and the agent reports progress
• One caption style: white text with a dark outline. Size (small, medium, large) and position (bottom, top) can change; font and colour cannot
• If your video already has captions in the picture, the new ones will sit on top of them
• No translation, no cutting, no highlight picking

Privacy

Your audio is sent to Deepgram for speech-to-text, and the transcript to DeepSeek for caption line breaks. Your video file itself stays in your workspace.

Want to see it first? Ask the agent to "run the sample".