Topic: Key decisions in long-form documentary editing

Question to the model: “What are the key editorial decisions when editing a feature-length documentary shot over two years with five different cameras and multiple audio sources, where the director is considering whether to use black-and-white, how long the film should be, and how much of the subject's interview to include?"

Model's response (summary): Consider the emotional tone when choosing between color and black-and-white. For length, follow the story — cut what doesn't serve the narrative. For interviews, identify the most informative and emotionally compelling statements and build the structure around them. Ensure visual consistency across cameras through color grading or matching.

My evaluation:
Correct: These are valid general principles. "Cut what doesn't serve the narrative" and "identify emotionally compelling statements" are not wrong. They are simply not decisions — they are descriptions of decisions. The model describes the destination without mapping the route.

Missed — 4 points where the real work happens:

1. Black and white is a structural decision, not an aesthetic one. The model frames the B&W choice as a question of "emotional tone" — as if it were a filter applied to a finished film. In practice, the decision to shoot or grade in black and white determines how you light, how you frame, and what visual contrasts you construct throughout production. In "Droga / The Way," the choice was made because the subject is fundamentally a person of contradictions: conservator and DJ, athlete and pianist, disciplined and emotionally fragile. Color would have softened these contrasts into a coherent surface. Black and white removes that surface and forces the viewer to read the person directly. The model would have said "black and white creates a timeless, serious mood." That is not why you make this choice.

2. Film length is a formal decision about density, not a result of cutting.The model says "follow the story — cut what doesn't serve the narrative." This assumes the length emerges naturally from the material. It doesn't. Length is a formal choice that precedes the cut. "Droga" was built upward from 10 minutes, not trimmed down from 40. The decision to aim for 20 minutes was a decision about the kind of film it would be: closer to poetry than prose, compression over explanation. At 40 minutes it would have become a chronicle — detailed, explanatory, linear. At 10 minutes it wouldn't have had room to breathe. 20 minutes was the space in which the film could be both spare and complete. The model has no framework for this kind of reasoning. It will tell you to cut until the story is served. It won't tell you what kind of story you're making.

3. Interview reduction is not about finding the best lines — it's about finding the irreducible ones. Two hours of interview. Three minutes in the final film. The model would approach this as selection: identify the most compelling statements, build structure around them. The actual process is closer to subtraction: start with silence, add only what the image cannot carry alone. What remained in "Droga" were not the most articulate or emotional moments — they were the moments without which the film's central idea (the journey matters more than the destination) would have been invisible or misread. "Good" material was removed not because it was weak, but because it pointed in a slightly different direction. This distinction — between good and necessary — is one the model cannot make, because it requires knowing what the film is about before you know what to keep. The editing was tested on viewers whose responses identified which silences held and which needed a single line to anchor them. That is an empirical process, not an editorial formula.

4. Multi-camera consistency is solved by constraint, not by grading. The model recommends "color grading or matching" to achieve visual consistency across five different cameras with different sensor characteristics and lenses. This is technically correct but practically incomplete. Grading can normalize exposure and color temperature. It cannot reconcile fundamentally different optical personalities — the rendering of a Blackmagic Cinema Camera and a Panasonic GH4 do not look the same after grading, and forcing them to often produces something that looks like neither. The solution used in "Droga" was upstream of post-production: cameras were not mixed within a single scene or shooting day. Each location had its own visual texture, consistent within itself. The black-and-white conversion then unified the material across sources by removing the dimension — color — where the differences were most visible. Lens differences, rather than being a problem to solve, became part of the visual language: different optical voices for different aspects of the subject's life. The model's answer treats this as a post-production problem. It is a production planning problem solved before the shoot.

Verdict: The model's response is professionally fluent and practically useless. It identifies the right categories of decision — tone, length, interview selection, visual consistency — without understanding how any of them are actually made. This is the most common failure mode in AI responses about documentary editing: correct vocabulary, missing reasoning. The gap between "cut what doesn't serve the narrative" and knowing what the narrative is, what serving it means, and what it costs to remove something good — that gap is where documentary editing actually lives. The model stands outside it.