On this page
How the AI reads your videos
Updated Oct 17, 2026
When a reply depends on what is in a video, the AI reads that video once and remembers what it learned.
What happens the first time
The first time a reply needs it, the video is read: its sound is transcribed, frames are looked at, and what was learned is saved as video context. Each version of a video is read once, however many comments come in.
If we can’t fetch the video
Upload the video file yourself, or write a few lines about it. Either way the AI has something to work from.
No sound in the video
If a video has no audio track, its video context says the sound was not heard. That tells you the AI only had the pictures and your notes.
What we keep
The downloaded video and its frames are deleted as soon as reading finishes, or within 24 hours if reading failed. We keep the video context, and you can view, correct or delete it.
What it costs
Reading shows up as its own AI line, at the model’s cost plus 50%. Before you read a video again you see an estimate. The free trial doesn’t read videos, so describe the video in your own words instead.