An AI editor and a professional human editor agreed on 84% of cuts when independently editing the same 4-hour video project, per a blind test posted by @kimmonismus. The result suggests AI can now replicate core editorial judgment on narrative structure, not just technical assembly.
Key facts
- AI and pro editor agreed on 84% of cuts
- 4-hour video project edited blind
- AI finished in under 90 minutes
- Human editors typically take 10–20 hours
- 16% disagreement involved pacing and emotional beats
In a blind test posted on X by @kimmonismus, an AI editing system and a professional human editor each cut the same 4-hour video project from scratch. Neither saw the other's timeline. The AI matched the pro on 84% of individual cuts — including trims, scene order, and transitions — according to the thread @kimmonismus.
Human editors typically spend 10–20 hours on a 4-hour project, factoring in review cycles and client feedback. The AI finished in under 90 minutes, @kimmonismus noted. The 16% disagreement mostly involved pacing choices and emotional beats — the pro sometimes held a reaction shot longer or cut to a close-up for dramatic effect, decisions the AI missed.
The test did not disclose the AI model used, the video genre (documentary, event, or narrative), or the editor's experience level. The sample size is a single project. Still, the 84% agreement rate is striking: it implies that a large fraction of editorial decisions — especially technical ones like removing dead air, tightening pauses, and matching shot reverse-shot — are now automatable.
What the delta tells us
The 16% disagreement is arguably more informative than the agreement. Professional video editing involves subjective judgment: when to break a rhythm, where to insert a reaction, how long to hold a silence. These choices define a cut's emotional arc. That the AI missed them suggests current models lack theory-of-mind reasoning about audience attention and narrative tension — a known limitation of transformer-based video understanding systems.
Industry implications
Consumer tools like Adobe Premiere Pro's Sensei, DaVinci Resolve's Magic Cut, and Runway's video editing features already automate rough cuts. But this test claims the AI can handle the fine cut — the stage where a human editor makes intentional, story-driven decisions. If replicable at scale, it could compress post-production timelines for YouTube creators, corporate video teams, and newsrooms, while shifting professional editors toward higher-level creative direction and client management.
Open questions
The source did not specify the AI model, training data, or whether the test was preregistered. Without a public benchmark or reproducible methodology, the result is anecdotal. A controlled study with multiple editors, multiple AI systems, and diverse video genres would be needed to establish generalizability.
Still, the 84% figure is within the range of human inter-editor agreement rates reported in film studies literature (typically 75–90% for structural decisions on raw footage). If confirmed, it would mean AI has reached parity with human editorial consensus on narrative assembly.
Key Takeaways
- AI editor matched pro on 84% of video cuts in blind test of 4-hour project.
- Suggests editorial judgment is partially automatable.
What to watch
Watch for a follow-up test with multiple editors and disclosed AI model (e.g., Runway Gen-3 or Adobe Sensei). Also watch for any peer-reviewed study replicating the 84% figure with controlled methodology and inter-rater reliability metrics.
