A generated rough cut and an approved production package are different deliverables. A useful comparison starts by defining the same finish line for both approaches. Otherwise a fast image test can be mistaken for a replacement for weeks of planning, approvals and finishing.
Status: This is a proposed comparison framework, not a completed client case study. It does not present a commissioned AI reconstruction, a timed test or measured cost savings.
Define the same deliverable
Choose one scene or short sequence. Record the script version, shot count, running time, aspect ratio, required performances and any product details that must remain exact. Define whether the output is a mood study, storyboard, animatic, shoot plan or final release. Do not compare one category with another.
Set review criteria before beginning. For example, a planning animatic might need to communicate camera positions and editorial rhythm; a release-ready commercial must also meet its product, sound, finishing and delivery requirements. Keep a written record of what the test includes and excludes.
Confirm the material can be used
Use footage and references that are authorized for the intended test. Establish permissions and release conditions for identifiable performers, music, logos and client material before generating or sharing alternatives. Do not treat an existing brand campaign as permission to publish a new version of it.
If a hypothetical brief is used, label it as hypothetical. If an older film is shown for context, credit the original team and separate that film from any new test. A public portfolio credit is not evidence that the original client commissioned or approved an AI experiment.
Record time without hiding work
Log the start and end of each stage: brief preparation, reference selection, generation or capture, selection, editing, revisions and final review. Distinguish elapsed calendar time from hands-on working time and render or queue time. Preserve unsuccessful attempts as part of the record.
File timestamps can help establish when an export existed. They do not, by themselves, account for planning, prior experiments, reused assets or approval delays. State those inputs when describing the result.
Compare total cost and accepted quality
Include the relevant labour, software, compute, licensing, production and finishing costs for each workflow. Show what was reused and which work was excluded. A generation charge alone is not a complete production budget.
Review both outputs against the agreed criteria. Record continuity failures, unusable shots, revision requests and compromises as well as successful frames. If the outputs serve different purposes or reach different quality levels, report that difference instead of a single improvement ratio.
What to publish with the result
A reviewable case study needs the approved brief, dated outputs, named responsibilities, an explanation of the method and the limitations. Show side-by-side material only when it can be shared with permission. Identify which shots are generated, captured or hybrid, and distinguish an internal test from a paid, approved campaign.
No percentage saving, time reduction or client endorsement is claimed here. Those statements should follow the completed record, not precede it. See what an AI director actually does for the creative workflow and AI direction for project inquiries.
Editorial correction, 9 September 2026: An earlier version described a Steve Aoki / Bud Light reconstruction with specific dates, budgets and timings. Supporting production records were not available for verification, so those assertions have been withdrawn. The existing URL is retained for continuity; this page does not establish that the described reconstruction occurred.