AI video editing analysis for YouTube

An AI video editing analysis watches and listens to your cut and tells you, with timestamps, where it is likely to lose viewers and what to change. Retti does this in Edit Review before you upload. Then it does what a read of the edit alone cannot: once the video is live, Video Review reads it against its real retention graph from YouTube Studio.

Updated November 2026 · By the Retti team

What an editing analysis should check

A transcript only shows the words, so an editing analysis has to watch and listen to the cut itself:

• Pacing. Where the edit drags or rushes, and where a section runs on after its point has landed.
• The opening. How long it takes to reach the first thing worth staying for, and whether the promise is clear in the first seconds.
• Audio. Silences that read as dead rather than tense, music changes that signal an ending too early, and level shifts.
• On-screen text and graphics. Whether a graphic or caption arrives with the point or after it.
• Energy. Whether the visuals, the music and the narration stay at the same intensity, or one drops away from the others.
• Rough cuts. Where a viewer would be missing context. Edit Review's voiceover mode finds those gaps and drafts the missing narration lines.

Every note should point at a second in the timeline and say what to change, so an editor can act on it without guessing.

Predicted or real: what each can tell you

Before you upload, nobody has watched the video. Any tool, Retti included, is reading the cut and estimating where people are likely to leave. That is useful, because the edit is the last point a retention problem is cheap to fix. But it is a read, not a measurement.

After you upload, YouTube Studio records when your real viewers left. That graph is the only measured answer. So Retti splits the job in two:

• Before upload: Edit Review reads the cut, Script Review reads the script and Hook Review reads the opening.
• After upload: Video Review takes the video's link and its retention graph, shows each real drop at the clock time it happened, and what was being said and how the section was built at that moment.

Comparing the two shows which of the things flagged before upload mattered to your viewers, and which did not.

What Edit Review gives you

• Notes pinned to the second they belong to, flagged by severity, so you know what to fix first.
• Your own timestamped comments next to the AI's, and your editor or team invited to the same review.
• Notes you can mark resolved, and a new version you can upload and compare after the fixes.
• A link you can share with someone who needs to see it.

Limits

• A read of the edit cannot know how your audience will react. Only the retention graph after you publish can.
• Retti is built for long-form YouTube videos, 8 minutes and up. It does not review Shorts.
• Video Review needs access to the channel's YouTube Studio for the graph, so it is for your own videos.

Examples

Retti publishes retention teardowns of real YouTube videos: the hook, the structure, the pacing and the moments a viewer is likely to leave, each at its timestamp. They are labelled as predicted reads, because Retti has no access to another channel's retention graph. On your own video, Video Review replaces that estimate with your real curve.

Related

• Edit Review: the tool, and reviewing with your editor or team.
• How to review your own edit before you publish.
• How to read a YouTube retention graph.

Frequently asked questions