What you're evaluating

MondoCut is our AI auto-editor — it takes raw user footage and produces a finished edit by itself. Your job is to judge how well MondoCut used the footage it was given — how it selected, arranged and edited that material into a compelling edit. For every clip: watch it and score each of the four areas 1–5 against the criteria shown, plus an emotion rating. Per-clip notes are optional. Then, at the end of each group, write a detailed category summary — the common issues and patterns across that whole video type. That category write-up is where the real analysis lives. The clips are split into groups — do one group at a time.

How to give feedback — read this first

• Don't critique the raw footage. The source clips are random user footage — some are shaky, poorly framed, or low quality. That is not what we're judging. Don't lower a score or write a comment because "the original shot is bad."
• Judge how MondoCut used the footage. Did it pick the best available moment and turn it into a compelling edit? If a clip feels weak, the issue is that the system chose or cut it poorly — not that the footage was captured poorly.
• Point toward the fix. Frame your feedback as suggestions to help the AI build a stronger edit: story, structure, pacing, clip selection, captions / VO, motion graphics, etc.
• A good edit is about story and feel — not technical precision. Cutting exactly on the beat for its own sake can feel robotic. Every cut should serve the story or the performance.

★ This round (0729) — add these two things to every scenario comment

Two extra asks from the team this round. Please include both in your written comments:

• What changed vs. the last version? In your comment for each scenario, say how this version differs from the previous round — what specifically got better or worse. This lets us pinpoint the exact change instead of guessing from the feedback.
• Timestamp specific problems. When you point out a specific issue, note the timestamp (e.g. "0:12 — cut feels abrupt") so we can run targeted analysis next round.

The 4 areas you're scoring

Scale: 1 = worst, 5 = best. Score each area on its own against its criteria — don't let one good or bad thing sway the rest.