Adjust subtitle timing with AI
Move a subtitle track forwards or backwards, convert a text file for a web player or produce a video with captions fixed to the picture.
To adjust subtitle timing with AI, first check whether the gap stays the same from one end of the programme to the other. A late first caption and an equally late last caption suggest a fixed offset. A gap that widens as the film plays calls for a different correction.
The subtitles skill works with an existing text track. You can move its timings, convert between SRT and WebVTT, extract the first supported embedded subtitle stream or burn captions into a fresh video. Decide whether the audience needs an on/off control before choosing the final format.
What it does
- Delay the whole track after adding a short title card to the opening.
- Convert an SRT file to WebVTT for a web player that accepts a separate caption track.
- Extract the first text subtitle stream from a container with an existing track.
- Burn approved captions into a copy that will always display them.
- Review long lines near visual details that the captions might obscure.
Install and use
- Install the skill with uv, Python 3.11 or later and FFmpeg. The burn step also uses Pillow, which uv can fetch.
- Ask, “These captions appear a second too early. Delay every cue by one second and save a separate subtitle file.”
- Test it with the original film at several points. For permanent captions, preview the rendered video at its intended viewing size.
FAQ
Source
Made by Timelapse Tech. Original skill: subtitles, published on GitHub under the MIT licence.
View on GitHub