Pre-Visualization Prototype Review (QA)

Pre-Visualization Prototype Review (QA)

1. 01

   Lock the client-approved script and shot list before any render is queued

   Freeze the storyboard, voiceover copy, and aspect ratio in a single signed-off doc so render iterations are judged against a fixed reference rather than drifting taste.

2. 02

   Cap the first render batch at three model variants per shot

   Run the same prompt through a multi-model workspace (Froging routes Kling, Veo, MiniMax, Wan, and Seedance) and stop at three; beyond that the review cost exceeds the marginal quality gain.

3. 03

   Screen every clip for lip-sync drift and hand or text artifacts at full resolution

   Watch at 100 percent zoom on the frames where a face speaks or a logo appears; MiniMax H3 clips carry native audio and lip-sync, so audio-visual mismatch is the most common rejection cause.

4. 04

   Verify duration and resolution match the delivery spec, not the tool default

   Short-form slots typically run 5 to 15 seconds; confirm 2K output where the placement demands it (AIProductAds renders 2K from prompts or product images) and re-render rather than upscale.

5. 05

   Score each approved clip against the brand kit: palette, type, logo safe area

   Reject renders that drift from the client palette even when the motion reads well; style drift is the failure mode clients notice first in a review deck.

6. 06

   Log rejected prompts and the reason code in a shared sheet

   Tag each rejection as artifact, drift, duration, or audio; after two projects the log becomes the agency's prompt library and cuts first-pass rejection rates.

7. 07

   Route approved prototypes to the human finishing brief, not straight to the client

   Attach the winning clip plus a one-line note on what human direction will fix (grade, sound design, motion graphics) so the upsell to premium finishing is priced before delivery.