In this guide
Create or correct a timed transcript, then upload it as a caption track or add it in a video editor. Check every line against the audio, include meaningful non-speech sounds for accessible captions, and review the final playback.
To add subtitles to a video, create timed text for the speech, then upload it as a caption track or place it over the video in an editor. Review the words, timing and readability against the original audio before publishing. Automatic captions are a starting point, not a finished transcript.
For more on CapCut recognition and caption-file export, see how to add captions in CapCut. For a TikTok-specific workflow, see how to add captions on TikTok. Choose the output first: a separate caption track can be turned on or off when the player supports it, while text burned into the picture remains visible in the exported video. You can provide both when your distribution workflow supports them.
Choose a caption workflow
| Your goal | Approach | Check |
|---|---|---|
| Viewers can turn captions on or off | Upload a timed caption file or create a track in the video platform. | Player support, language, timing and the published track. |
| Text always visible in the video | Add timed text in a video editor and export the picture with it. | Contrast, placement and whether text covers important visuals. |
| Translate spoken dialogue | Prepare and review a separate timed translation. | Meaning, language label and timing against the scene. |
Terminology varies by region, but accessible captions convey speech plus relevant non-speech audio, such as a doorbell or an off-screen speaker. W3C's caption guidance explains this distinction and why automatic text needs human review.
1. Prepare and check the words
- Start with the clearest available audio. Listen to the complete video and note speakers, names, numbers and meaningful sounds. If you have original dialogue tracks, use them as a reference.
- Create a transcript. Type it yourself or use a transcription tool that suits your privacy and language needs. Check its output against the recording, especially names, technical terms, overlapping speech and negations.
- Divide it into readable cues. Break the text at natural phrases. Give viewers enough time to read each cue without leaving it on screen after the relevant speech has ended.
- Add meaningful non-speech information. Where needed, identify an unseen speaker or describe a sound such as [doorbell rings]. Keep the description relevant to understanding the scene.
Saturalabs does not transcribe speech or generate subtitle files. Its video-to-audio converter can prepare the mixed soundtrack from a supported local MP4 or MOV up to five minutes and 50 MiB, subject to browser codec, channel and 50 MiB output limits. It does not separate speech from music. The transcription guide compares recording preparation and verification.
2. Make a timed caption file
A common format is SRT: each cue has a number, start and end timestamps, text, and a blank line before the next cue. This example describes a fictional clip; replace the words and timings with your own:
1
00:00:01,000 --> 00:00:03,200
Welcome to the studio.
2
00:00:03,800 --> 00:00:05,100
[doorbell rings]The first cue begins at one second and ends at 3.2 seconds. Use a plain-text editor or caption editor, save as an .srt file, and verify the target platform accepts that format. Do not assume a transcript without timestamps is already synchronized. YouTube lists accepted formats in its subtitle-file documentation.
3. Add captions to a YouTube video
- Sign in to YouTube Studio, select Subtitles, and choose the video.
- Select Add language, choose the language and select Add under Subtitles.
- Upload a supported timed file, use Auto-sync with a transcript, or type captions manually. YouTube's caption instructions describe the available options and requirements.
- Review the words and timing in the editor and publish the track. YouTube's editing guide explains how to adjust timestamps or duplicate and edit an automatic track.
- Play the published video with captions enabled. Check that the intended language appears and that cues remain synchronized throughout.
Auto-sync uses speech recognition and may take time; Google notes language and audio-quality limitations. A generated caption can mishear a name or change a sentence's meaning, so compare it with the actual audio rather than publishing it unchecked.
4. Burn subtitles into a video in CapCut
- In CapCut Desktop, create a project, import your video and place it on the timeline.
- Open Text → Auto captions, select the spoken language and choose Create. CapCut's desktop subtitle guide describes these controls; layouts and feature availability can vary.
- Review and correct each caption on the timeline. Add missing speaker information or meaningful sounds where appropriate, and adjust start and end times.
- Choose readable text styling and placement, then export the video and review the actual saved file. Text rendered into the picture cannot be switched off like a separate player caption track.
If the language or recognition feature is unavailable, create timed text manually in your editor or use a separate caption workflow. Check the service's privacy and upload terms before sending sensitive recordings to any online transcription tool.
Review the final captions
- Accuracy: Listen for misheard names, numbers, omissions and changes in meaning. Avoid guessing at unclear speech.
- Timing: Check the beginning and end of each cue, especially at speaker changes, pauses and fast dialogue.
- Readability: Use clear contrast and a readable size. Keep captions away from important on-screen text and interface overlays.
- Audio context: Include relevant sound effects, music cues and speaker identification where needed to understand the content.
- Final playback: Watch the exported video or published caption track on the intended player and a small screen. If you revise the video edit, recheck sync.
If background audio makes a word hard to understand, compare the original track with any cleaner source. Saturalabs' voice isolator can test a supported mixed recording but can leave bleed or alter words; AI downloads require an account and sufficient credits, with cost shown before download. A cleaner-sounding result does not prove an unclear word was recovered.
Common questions
Are captions and subtitles the same?
The terms are sometimes used interchangeably. For accessibility, captions should communicate the speech and relevant non-speech audio needed to understand the video; a dialogue-only translation may omit that context.
Can viewers turn off burned-in subtitles?
No. Burned-in text is part of the picture. Use a separate caption track when the platform and player support viewer-controlled captions.
Does an SRT file contain the video?
No. It contains timed text, not the video or audio. Attach or upload it alongside the correct video and verify timing after any edit.
Can Saturalabs automatically create subtitles?
No. Its tools can help prepare or isolate audio for review; a separate transcription or caption editor is needed to create and time the text.