Auto caption generator for ads
Genyad burns word-synced captions into every variation automatically, using the timing data its voiceover synthesis returns rather than a separate transcription pass. Captions are styled per variation and kept clear of each placement's UI, which makes caption treatment itself something you can test.
Last updated
How to do it in Genyad
- Generate a variation. Captions are produced automatically from the voiceover timings.
- Pick a caption style from the typography catalogue, or leave the agent's choice.
- Check the safe zone in the preview for the placement you are exporting.
- Edit caption text if a line is too long, and the sync is preserved.
- Export per placement. Each ratio gets its own caption layout.
What good looks like
- Two to four words on screen at a time for vertical placements. Full sentences do not get read.
- Contrast that survives a bright background, which usually means a stroke or a plate behind the text.
- Nothing in the top 14% or bottom 20% of a 9:16 frame.
- Captions that carry the argument on their own, because a large share of impressions are muted.
Mistakes that cost you the test
- Relying on platform auto-captions. They are optional, often off, and styled by the platform rather than by you.
- Line-level captions on vertical video. Word-level pacing reads far better at phone size.
- Small text. If it is not legible in a thumbnail-sized preview it is not legible in the feed.
- Captions that duplicate the voiceover exactly at length. Trim for reading speed, not for transcript accuracy.
Frequently asked questions
Are captions added automatically?
Yes, on every variation, generated from the voiceover's word-level timing rather than from a separate transcription step.
Can I restyle captions?
Yes, per variation, from a typography catalogue. Caption treatment is worth testing like any other creative variable.
Do captions move for different aspect ratios?
Yes. Each export is laid out for that placement's safe zone rather than scaled from one master.
Can I edit caption text?
Yes, and the sync is preserved because timings come from the audio, not from the text length.