https://reddit.com/link/1xf8if8/video/stnp2qrmfnxh1/player

I’m building a tool that turns uploaded videos into short recaps with narration and subtitles.

A script can sound fine on its own and still produce a confusing video. The narration describes an action, but the selected footage doesn’t show it.

I’ve been working on that gap:

Grounding the script in the source transcript.

Writing narration that can be matched to footage that actually exists.

Checking scene selections before rendering.

Keeping the voice timing and volume consistent.

One particularly strange issue: the AI narrator could spontaneously laugh. That needed fixing too.

The attached one-minute video is a sample recap.…