How to create and edit AI captions for video

1. Generate the transcript

Open a recorded video in the library and select Captions. Choose Generate Captions to transcribe the clip’s audio on your device.

The first use downloads a speech model. This can take time and use several hundred megabytes of data; speed depends on your computer and clip length. Once cached, the model can be reused, subject to browser storage.

Local transcription does not need a personal API key or managed credits. This is the step that listens to the audio and produces words with timings.

2. Review and edit the words

Read the generated segments and replay uncertain moments. Correct names, numbers, specialist terms, and punctuation. Delete unwanted caption segments if needed.

Recognition is multilingual, but accuracy differs across languages and accents. Mixed-language speech, background music, and overlapping voices deserve extra attention. Keep the original meaning when correcting text.

3. Choose optional AI text help

Improve With AI is a separate text step. It sends caption text for refinement, not another pass at listening to your audio. Check the suggested corrections.

In AI Settings, choose a supported provider or compatible HTTPS endpoint and model. The integration expects a compatible chat-completions API; arbitrary service keys are not interchangeable. Provider charges and terms apply.

Personal keys use session storage, so a later browser session may require entering the key again. Managed AI offers email sign-in and credits when enabled for your account. Prices are shown before purchase.

4. Replace subtitles already in the source

Baked-in subtitles are part of the picture. When a vertical crop cuts that line, adding new words alone leaves two visible captions.

In the updated editor, seek to a moment showing the old subtitle. Enable Replace subtitles already in this video. The alignment view pauses the clip and magnifies the lower portion so you can position the guide.

Move and resize the strip over all the old words. Choose a color and set opacity to 100% for a complete cover. Lower opacity allows the picture and text underneath to show through.

Select Preview final to check the full frame. New captions are centered in the strip. The tool covers pixels; it does not reconstruct the background. Check several moments if the old subtitles move or change height.

This feature may require a newer extension package than the version currently available in the store.

5. Choose an export

Save Transcript keeps editable timed captions with the recording. Reopen the track to revise the words.

Export SRT creates a common subtitle file for compatible editors and platforms. Export VTT creates WebVTT for compatible web players. Subtitle files contain text and timing; they do not include the visual cover strip.

Burn Captions Into Video renders the words and any enabled cover into the picture. This keeps captions visible wherever the file plays. Burned-in words are not a toggleable subtitle track, so check the result carefully.

6. Prepare a title and social caption

Optional AI assistance can draft a title and social caption from the transcript. Edit them to match the content and audience. Check claims and hashtags before sharing.

Use the title for the recording, save the social text, or copy it for your publishing workflow. The extension prepares material; it does not publish on your behalf.