Skip to content

Auto caption generator

The words spoken in your video written as captions, the current word highlighted as it is said, ready for Reels, TikTok and Shorts.

The auto caption generator with a video loaded: the spoken language, the caption style, the words per caption and their position

Guide

How to add captions to a video automatically

Why captions matter

Most people scroll through Reels, TikTok and Shorts with the sound off, in a queue, on a bus or late at night. Without captions they see someone talking and swipe away. Captions also help viewers who are deaf or hard of hearing, and those watching in a language that is not their first.

Typing captions by hand and timing each line takes longer than recording the video. This tool writes them from your speech, highlights each word as it is said in the style short videos use, and lets you check the text before the video is made.

How to use it

  1. Sign in with your free account and drop your video on the box above.
  2. Choose the spoken language, or leave it on automatic, and pick a style, the words per caption and the position.
  3. Press Write the captions: the page takes out the sound and our server writes down the words.
  4. Read the captions in the list and fix any word, then press Make the video.
  5. Watch the result and download the video, and the SRT or VTT file if you need it.

How it works

Your browser takes the sound out of the video and turns it into a small mono file at 16 kHz, about 2 MB a minute, which is all that is sent. On our server Whisper, an open-source speech recognition model, writes down each word with the moment it starts and ends. The sound is used only for that and is not kept.

Back in your browser the words are grouped into captions by the same rules as our AI video maker, and its engine draws them over every frame. The video is made on your device, so it is never uploaded, and its sound is copied into the new file unchanged.

Choosing the style

  • Impact: large capital letters with a green highlight, the most common look on TikTok.
  • Classic: clean lettering with a soft shadow, readable on any video.
  • Box: a solid band behind the words, best over busy or bright footage.
  • Karaoke: a coloured box moves from word to word, good for fast speech.
  • Words per caption: 1 for a punchy one-word rhythm, 3 for most short videos, 6 for calmer explanations.

Limits to know

Speech recognition is very good with clear speech and can still mishear names, brands, slang and words drowned by music, which is why the review step comes before the video is made. Cleaning the voice first with the background noise remover improves the result on noisy recordings.

Each account has 3 captioned videos a day, plus 2 for each friend who joins through your invite link, because each one uses our server. Videos can be up to 15 minutes long; on a phone, up to 5 minutes and 150 MB.

Before you post

Captions work best on a tight video: the silence remover takes out the pauses first, so the captions keep moving. If your video is horizontal, the vertical video converter puts it on a 9:16 frame before you add the captions. And the AI video maker creates a complete captioned video from a single sentence when you have no footage at all.

Questions about automatic captions

How are the captions written?
Only the sound of your video is sent to our server, as a small audio file, where Whisper, an open-source speech recognition model, writes down each word with the moment it starts and ends. The words come back to your browser, which draws them on the video.
Is my video uploaded?
No. The video stays on your device: the page takes out its sound, sends only that, and draws the captions on the video in your browser. The sound is used to write the words and is not kept on our server.
Why do I need an account?
Writing captions uses our server for every video, so each person has 3 captioned videos a day, plus 2 more for each friend who joins through your invite link. The account, free with Google or an e-mail address, is how those videos are counted.
Can I fix a word the tool got wrong?
Yes. Before the video is made, every caption appears in a list you can edit. Change a word, add punctuation or delete a filler, and the words of that caption are timed again across the same moment, so the highlight still follows the voice.
Which languages does it understand?
It detects the language by itself, and you can also pick one of 18 common languages, including English, Portuguese, Spanish, French, German, Hindi, Arabic and Japanese. Picking the language helps with short videos and strong accents.
What do the caption styles look like?
They are the styles of our AI video maker: Impact, in large capital letters with a coloured highlight; Classic, clean with a soft shadow; Box, with a solid band behind the words; and Karaoke, where a coloured box moves from word to word.
Can I download the captions as a file?
Yes. Along with the captioned video you get SRT and VTT files with the same text and times, to upload to YouTube or to any editor that reads subtitle files.
How long can my video be?
Up to 15 minutes. On a phone the page accepts up to 150 MB and 5 minutes, because the video is made in the phone memory. A short video for social media takes well under a minute to caption.

Last updated:

Was this page helpful?