Automatic captions: add words to short videos
Automatic captions turn the speech in a video into timed text. Start with the caption tool in the editor or publishing platform you already use, generate a draft, then check the words and timing before sharing. If you are making short clips in Reel Studio, its Captions button can make a separate copy with the spoken words burned into the picture. The useful shortcut is getting a first transcript automatically. The final listen still belongs to you.
This guide takes a short spoken video from finished audio to checked captions. It also explains which kind of caption file you need, when a manual pass is enough, and where Reel Studio’s clip-captioning workflow fits. The YouTube and CapCut features linked below were checked against their own help pages as of November 2026.
Choose what the viewer should receive
Before generating anything, decide how the words need to reach the viewer. A caption track and text painted onto the video solve different delivery problems.
| Delivery | What the viewer receives | Choose it when |
|---|---|---|
| Closed-caption track | A separate timed text track that the player can show | Your destination accepts captions alongside the video |
| Burned-in captions | Words permanently visible in the video picture | You want the words present wherever the file is played |
| Both versions | A clean video with a track, plus a captioned copy | You deliver the same piece to different destinations |
Burned-in words cannot be switched off or restyled by the viewer. A separate track gives the player control, but you must check that it accompanies the video at the destination. Keep a clean original so you can change direction later.
Also decide what the captions must communicate. A short promotional clip may use one word at a time for emphasis. An interview may need complete phrases, speaker identification and meaningful sounds. A decorative word overlay alone does not meet every viewer’s access needs. Choose the presentation for the audience and the delivery requirements.
Make automatic captions in the tool you already use
If your video already lives in a working project, start there. Moving it to another service introduces another export and another version to keep straight.
If the video is going to YouTube
As of November 2026, YouTube’s automatic-captioning help says it can generate captions for long-form videos and Shorts. Its review route is YouTube Studio, then Subtitles, then the video and its caption track. Automatic captions may arrive after upload, so check again before treating an absent track as a failure.
YouTube’s instruction is clear:
“Always review automatic captions”
Review the track and correct recognition errors. If you have an approved transcript, its caption-creation guide also describes uploading a file, using Auto-sync, and typing manually. For a video that only needs a track on that platform, this is a sensible place to finish.
If you are already editing in CapCut
As of November 2026, CapCut’s own online subtitle guide describes this sequence: upload the video, put it on the timeline, open Captions, choose Auto captions, select the spoken language and generate. Review the text before exporting the video.
The same page describes adjustments to text size, position and colour, plus separate SRT export. Stay in that editor if you need to adjust caption wording, line breaks and styling alongside the rest of your edit. Check the options available in your account before promising a particular delivery format to a client.
Whichever route you take, generate captions after the spoken edit is settled. Removing a sentence afterward means its captions and everything following it need another timing check.
Prepare the audio before you transcribe it
Listen once with the picture out of view. Can you understand the sentence, the product name and the closing instruction? If you cannot, captions need more than a quick approval.
For a new recording, reduce avoidable background noise and leave space between speakers. For an existing clip, choose the version with the clearest voice. Keep music low enough that the spoken words remain distinct. These are preparation choices, without any promise that a recogniser will get every word right.
Write down the spellings a general transcript is likely to need help with:
- A person’s name and how it should be capitalised.
- The exact product or business name.
- Numbers, measurements and dates that must survive unchanged.
- A technical term or phrase in another language.
- A small word that changes the meaning, such as “can” or “can’t”.
Keep that list beside the video during review. For an ad, check the offer against the approved copy as well as against what the voice actually says. A polished caption of a mistaken price is still a mistaken price.
Caption a short clip in Reel Studio
Reel Studio is useful when the clip is already in your bank, or when you want a simple word-by-word copy of a short video. You can upload your own video to the bank as well as use a take you made in the studio.
For a standalone clip whose words must be read from its audio, the limit is 30 seconds, and the clip needs sound. The current style shows one bold yellow word at a time. Choose another caption editor if you need paragraph-style subtitles, extensive text corrections or a separate subtitle file. This workflow makes a captioned video copy.
- Open the studio and sign in if asked. Choose the video in your bank, or upload it there first.
- Open the clip on the stage. Use the uncaptioned original so you have a clean starting point.
- Press Captions. The button first says it is reading the words, then shows progress as it makes the copy.
- Keep the studio open while it works. The copy is assembled in your browser. You can press the active button again to stop; a stopped run keeps no copy.
- Play the new captioned clip. It appears beside the original in your bank. Check every spoken word and the moments when it appears.
- Download the checked copy. Keep the original too, especially if another destination later needs different captions.
Captioning the clip does not spend generation credits. Making new images, voices or video takes is a separate action with its own credit price.
If Captions is disabled, read its explanation. A silent clip has no speech to read, a captioned copy already has its words, and an assembled cut is captioned through its film instead. For a UGC ad cut, turn on its Captions switch and assemble it again. The UGC ad guide explains the wider creation workflow.
Check the words first, then the timing
Review in two passes. During the first, listen and read together. During the second, concentrate on when the words arrive and whether they remain readable against the picture.
| Check | What to look for | What to do |
|---|---|---|
| Meaning | A missing negative or changed number | Correct it in an editor that supports caption-text editing |
| Names | A brand written as ordinary words | Use the approved spelling |
| Timing | Words arrive before the voice or linger into another shot | Adjust the timing and replay the transition |
| Placement | Text covers a face, a product label or another title | Reposition it in a suitable editor |
| Delivery | The uploaded video has duplicate or missing captions | Check the final destination’s caption settings |
For example, imagine a clip saying, “It fits two bottles, not three.” A transcript that drops “not” reverses the message. An animated caption treatment can make that mistake more prominent because the viewer’s eye follows each word. Read for meaning before choosing effects.
In a scripted Reel Studio UGC beat, captions can use the beat’s written line with timing taken from the audio. If the sound does not match the line, timing can be estimated, and the studio says so. Treat that notice as a reason to listen again. Compare the displayed script with the actual spoken line: intended text and recorded speech can differ.
If an ordinary uploaded clip is transcribed incorrectly, use a caption editor with text controls and the clean original. Repeatedly burning the same incorrect transcript into another copy will not repair the words.
Keep captions readable on a phone
Watch the exported file at the size someone will actually see it. Text that seems generous on a large monitor may be hard to follow in a small feed preview.
Start with three checks: contrast against the busiest frame, separation from important visual details, and room around the edges. Then preview the video in its intended destination. Buttons, account names and descriptions can occupy the same space as captions that looked clear in a plain video player.
Watch once with sound and once muted. With sound, you can catch timing errors. Muted, you notice missing context and places where the text rushes past. Ask whether a person can follow the thought, including who is speaking and any sound that changes the meaning.
Keep the clean master, the captioned export and any editable caption file clearly named. When the spoken edit changes, return to the project or clean master and check the whole affected section again. Painting a second caption layer over an old one creates a version you will have trouble repairing.
When the manual way is enough
For a single short line, typing the words yourself may be the easiest route. It also gives you direct control when a proper noun or an unusual pronunciation matters more than getting a draft quickly.
Manual captioning makes sense when:
- There are only a few words and you already know the exact wording.
- The recording is too unclear to trust an automatic first pass.
- You need deliberate phrase breaks or detailed descriptions of important sounds.
- The destination requires a specific caption format that your current tool cannot deliver.
For repeated spoken clips, automatic transcription can remove the initial typing step. Keep the same review checklist for every export rather than assuming the next clip will behave like the last.
If you already edit in CapCut or publish only to YouTube, try their existing caption workflow first. If your next job is a short clip in Reel Studio, open it in the studio, make a captioned copy, and judge the words against the actual audio before you share it.