Should you fact-check an AI agent's lens and aspect ratio claims before locking your visual direction?

Yes — verify lens class, aspect ratio, and lighting-source claims before locking your visual direction. In one documented production, an AI agent labeled The Conjuring 'anamorphic'; when challenged, it corrected to spherical, 35mm, 2.40:1 hard matte. Spherical means circular bokeh and no horizontal lens flares — an error that would otherwise shape every downstream frame.

Challenge the technical claims before any of them get written into your treatment or locked context — lens class, aspect ratio, and lighting source are the three to verify first. invideo is an agentic video creation tool where the invideo agent analyzes reference films and stores its findings in persistent context, which is exactly why a factual error at this stage compounds: once locked, it isn't re-checked.

The documented case shows both the failure and the fix. During a horror short built on a James Wan director protocol, the invideo agent's analysis listed 'anamorphic' for The Conjuring. The director pushed back, and the invideo agent self-corrected: Wan shoots spherical, 35mm, with a 2.40:1 hard matte — widescreen by extraction, not optics. That distinction is material, not cosmetic. Spherical glass produces circular bokeh and no horizontal lens flares; prompt for anamorphic and you bake oval bokeh and streak flares into a style that never had them.

Run the same challenge on each claim type. For lens: ask the invideo agent how the lens choice manifests on screen (bokeh shape, flare behavior, distortion) and whether its source analysis supports that. For aspect ratio: ask whether the ratio comes from the optics or from a matte — confusing the two changes the prompt language. For lighting: demand source attribution, not adjectives — specifying 'warm yellow from the lamps only, like all the refs' produces more accurate results than generic 'warm lighting', and the same Wan protocol stored a specific 85:15 dark-to-light ratio as quotable prompt grammar. Cross-check the answers against independent references such as IMDb technical specs or American Society of Cinematographers material before you lock.

The payoff is in propagation math. Lens and aspect ratio sit as a fixed element in a structured prompt assembly — one documented pipeline enforced a 9-element order across every frame, and another output 12 parameters per shot including lens and lighting plan. The horror short that surfaced the anamorphic error ran roughly 400 video generations and 30 image generations over 2 days; an uncorrected lens class would have silently steered all of them. Once the facts check out, you can go one step further and stress-test the whole document by asking the invideo agent to apply the style to a genre the director never worked in — coherent output confirms it internalized grammar, not surface aesthetics.

Watch some of these to see what works for you:

Good catch. Wan shoots spherical, not anamorphic. The Conjuring: 35mm, 2.40:1 hard matte. Widescreen by extraction, not optics. I had 'anamorphic' in my earlier analysis. I'll correct it.

— the invideo agent, self-correcting when challenged during a documented James Wan-style horror short production