Free video caption generator

How to add captions to a video

There are two ways to put captions on a video. A caption file (SRT or VTT) is uploaded next to the video, and the platform draws the text: viewers can turn it off, and search engines can read it. Burned-in captions are part of the picture, so they show everywhere, in your font and colors, with the sound off.

For short-form UGC, burn them in: most people watch TikTok and Reels with the sound low, and a viewer who can't read the hook scrolls on. Keep the SRT anyway: it's the transcript a brand may ask for, and the text you'll reuse for the post caption.

SRT vs VTT: which caption file do you need?

Both are plain text: a start and end time for each caption, then its lines. The differences are small but matter to the software that reads them.

1
00:00:01,200 --> 00:00:03,050
I tried this serum
for 30 days

2
00:00:03,050 --> 00:00:04,400
and here's what happened

WEBVTT

00:00:01.200 --> 00:00:03.050
I tried this serum
for 30 days

00:00:03.050 --> 00:00:04.400
and here's what happened

A caption file has no font, color or word highlight: the player decides how it looks. The styles in the preview above show how the same timing reads when burned in.

Caption style tips for UGC videos

• Short captions, read in a glance. One or two lines, three to six words each. The defaults here are six words and 24 characters per line; for fast talkers, try three words or One word.
• Keep them out from under the app. The bottom third of a TikTok or Reel is covered by the caption, the username and the buttons. The dashed line in the preview is each app's published safe zone; the TikTok safe-zone guide and Reels safe zone pages have the exact margins.
• Make the spoken word stand out. Word-timed styles (a highlighted or boxed word that moves with the voice) keep eyes on the caption, which is why they're everywhere in UGC ads. Captions here carry word timings so you can see it in the preview.
• Caption the hook first. The opening line does the work; check that its caption is on screen from the first frame and spelled exactly right.
• Check every name, price and claim. Speech recognition guesses at brand names and numbers, and a brand will notice. Fix them in the list before you download.
• High contrast, no clutter. White text with a shadow or a solid tag reads on any background; skip outlines in three colors.

The UGCLive app does all of this inside the edit: it captions every word as you cut, with the same Highlight, Tag, Karaoke and One word styles, and burns them into the export. Writing the script first? The UGC script template and the teleprompter help you say it in one take.

What runs on your device

The page reads the sound from your file with the browser's own audio decoder, then runs OpenAI's Whisper (MIT license) through Transformers.js (Apache 2.0) and ONNX Runtime Web (MIT), in a background thread. It uses your graphics chip through WebGPU where the browser has it, and the CPU otherwise. Your video and its transcript stay in this tab; close it and they're gone.

Clips can be up to 10 minutes and 1 GB; longer ones are slow in a browser and can run out of memory on a phone. The audio is transcribed in parts of up to 30 seconds, cut at quiet moments, and silent parts are skipped, since Whisper tends to invent words in silence.