Tom and Jerry | Live Action: How the Painted Tunnel Gag Was Made

How do you keep the cat and mouse consistent across 13 shots?

Each animal gets its own character model sheet, and the video prompt treats those sheets as the identity source. Both sheets were generated with GPT Image 2.5 Flare at 2K in a 4:3 frame, showing the same animal in front and side head views, full-body poses and labeled color swatches for fur, eyes and other details. The video prompt then restates the key traits in words (tabby stripes, white chest and green eyes for Milo; round ears and a long bare tail for Finn) and asks the model to match each image. Behavior rules live in the character description too: the cat only stands upright while it paints, and the mouse only sprints low to the ground.

How was the painted tunnel gag written for the AI video model?

The whole gag is a single Seedance 2.5 generation: a 30-second, 16:9 film at 480p written as 13 numbered shots joined by 12 hard cuts, with one shot left uninterrupted. A global block sets the look once (low warm late-afternoon sun, long shadows, dust in the air, weathered granite and smooth tracking moves) so every shot shares the same light. Scale is spelled out explicitly, with the tunnel at roughly cat height and the boulder big enough to hide behind, which keeps the props believable next to real-looking animals. The prompt asks for the absurd moment to be played completely straight, and that deadpan contrast is what makes the mouse slipping into the painting and the cat's failed follow-up read as slapstick.

What is inside the template and how do I remake it with my own characters?

The canvas holds five reference image nodes, including character sheets for the cat and mouse and empty scene plates for the painted rock face and the gravel road with its boulder, plus one video node that ties them together. To remake it, regenerate the sheets with your own pair of characters while keeping the same layout, or swap the plates for a different location. Then edit names, traits and actions in the shot list while leaving the global look block and shot structure intact. Generate, trim the result on the timeline and export to mp4 or FCPXML.

How to remake Tom and Jerry | Live Action in Filmera

1. Duplicate the template — Open the template page and duplicate it to get your own copy of the canvas with all reference and video nodes.
2. Swap the character sheets — Regenerate the two character reference nodes with GPT Image 2.5 Flare using your own characters, keeping the model-sheet layout with front, side and full-body views.
3. Replace the scene plates — Keep or regenerate the empty background plates for the illusion wall and the road, so the location matches your new story.
4. Rewrite the shot list — In the video node, update names, traits, scale notes and actions per shot, keeping the global look block and the hard-cut structure.
5. Generate with Seedance 2.5 — Run the video node and compare the result against your reference sheets, regenerating any shot where a character drifts.
6. Edit and export — Arrange and trim the clip on the timeline, then export to mp4 or FCPXML for further editing.