Original examples · Captions · public alpha
SRT vs WebVTT vs burned-in captions: examples and fixes
Test notes updated October 10, 2026 · by thecreatingco
Choose SRT for a simple timed subtitle file accepted by your destination editor, WebVTT for a browser text track, or burned-in captions when the words must be part of the video image. Keep an editable text version even if you burn the final captions into a video.
One example, two subtitle formats
These original files contain the same three manually written cues. They are timing exercises, not speech-recognition output. Play the 12-second silent test video below to see the WebVTT track. Its plain background intentionally keeps the words easy to inspect.
Download SRT · Download WebVTT · Download the silent test video
1
00:00:00,500 --> 00:00:03,000
A small idea starts here.The SRT cue above uses a comma before milliseconds. The corresponding WebVTT file begins with WEBVTT and uses 00:00:00.500. Changing a filename extension alone does not convert a subtitle file.
SRT vs WebVTT vs burned-in captions
| Delivery | Useful when | Tradeoff |
|---|---|---|
| SRT | A destination editor or upload service asks for SRT | Basic text and timing; do not expect the app’s visual caption style to transfer unchanged. |
| WebVTT | You are adding a selectable text track to an HTML video | The player and its caption settings affect presentation. |
| Burned-in video | The words need to appear as part of the image | Viewers cannot switch off those pixels; corrections need a new video export. |
A review workflow in Captions
- Open a short audio or video clip in Captions and run speech recognition. Start with material you can check against the original recording.
- Correct words before styling. Pay special attention to names, numbers, technical language and lyrics.
- Listen through each cue. Shift its start and end to match the intended speech, and split a long cue at a natural phrase boundary.
- Export the format your destination accepts. Re-import or play that actual downloaded file; a correct editor preview does not verify the export.
- For burned-in WebM, keep the tab visible while encoding and check the beginning, middle and end of the result.
Fix missing, drifting or duplicated subtitles
Nothing appears: turn captions on in the player, confirm the track language, and check that the cue times fall inside the video’s duration. The example starts at 0.5 seconds, so an empty first half-second is expected.
Every cue is early by the same amount: compare a clear word near the start and another near the end. A constant offset points to a shared timing shift; fix a copy and review again.
The error grows toward the end: check whether the video was trimmed, sped up or replaced after transcription. A single offset cannot fix a speed mismatch.
Two copies of the words: disable the external track briefly. If words remain, the video may already contain burned-in captions. Deliver one intended presentation rather than layering both.
WebVTT is UTF-8 timed text associated with a media element through a track. This demonstration uses three ordered cues with end times later than their starts. Source: MDN WebVTT format documentation.
Caption files here are hand-authored. We have not benchmarked Captions’ transcription accuracy or qualified its encoder across devices. Speech recognition and browser export remain alpha workflows requiring review.
Reuse the examples
These are original learning fixtures. You may reuse and modify them, including commercially. Read the fixture notes and limits. Keep originals and inspect actual exports before delivery.