AI Prototyping, Projects and Digital Art

NightCut grew out of a problem Dima had himself: he's a YouTube vlogger with a huge backlog of unedited travel vlogs, so he built himself the editing assistant he needed – running entirely on his own machine.

At its core sits the analysis: 52 analytical modules run over every single clip, looking at visual content, motion, audio, and quality. It doesn't stop at individual clips either – NightCut also understands the footage across clips, spotting repeats and retakes and automatically filtering out shaky, blurry, or otherwise unusable footage, so only the best takes make it into the final assembly.

That analysis is heavy lifting and runs overnight. By the next morning, every clip has a summary, a transcript, and a rating of its usable moments – and from there you can generate as many timelines as you like, on the fly, depending on what you need: a travel vlog, a highlights reel, or something else entirely.

The algorithms are still being refined, with a commercial product planned further down the line.

A full-scale startup product: users place pins on shared maps curated by interest groups. Built with developers from Ukraine, approximately 1,300 hours of development work.