Northline: A 73-Second Seedance 2.0 Brand Film Under $24

Project Type: AI brand film and interactive concept site (self-initiated)

Brand: Northline Practice Growth, a fictional concept for independent dental implant practices

Role: Creative direction, generation pipeline, motion continuity, and frontend build

Duration: July 20 to July 23, 2026

Northline Practice Growth is a fictional consulting brand I made to test a narrow production question: how much of a coherent brand film and interactive site can one person build with the current image and video models?

I chose independent dental implant practices because I already work in that category. The story is demand moving through one practice. A copper path enters from the lower left, tangles at intake, runs through consultation and treatment, and reaches a north marker after the team repairs the constraint.

This is a concept build. There is no client and there are no business-result claims.

Six independently generated scenes will drift unless the geography, people, materials, and visual metaphor are fixed before the first prompt runs.

A sequence of good video clips still reads as eleven separate clips when the camera jumps at every boundary.

A finished film does not become an interactive site by itself. Scroll position, seek behavior, mobile crops, loading states, and reduced motion still need a working system.

Cheap generation is an empty claim without the receipts and the rejected work included.

The continuity contract came first: one clay miniature practice, porcelain walls, a midnight-blue roof ribbon, five recurring people, and one copper pathway. The path had to survive every scene because it carries the business story without a voiceover.

Twelve 4K GPT Image 2 generations produced the six final scenes. The first set established the direction. The second set tightened the shared geography and kept the same

people, palette, and copper path in view.

The six scenes are signal, diagnosis, system, patient path, rhythm, and future. Each became an eight-second Seedance 2.0 dive. The camera moves into the current constraint, follows the repaired path, and returns to the full practice after the team can repeat the system.

The sequence follows the operating work in order: see the whole practice, find the constraint,

repair it, and make the new system repeatable.

Five additional Seedance clips connect the six dives. Each connector begins on the actual last frame of the preceding clip and ends on the actual first frame of the next one. The camera can cross a wall or rise back to the master view without cutting to a different version of the world.

The final chain has eleven clips and 73 generated seconds. The stitched web encode runs 73.458333

seconds at 1920 x 1080 and 24 frames per second.

The site loads each clip into a seekable browser blob and maps scroll position to video time. Scroll down and the camera moves forward. Scroll up and it runs backward. Copy and navigation change with the active scene.

The engine is plain JavaScript with no runtime dependency. On a smaller screen it changes the focal crop. With reduced motion enabled, it leaves the stills in place and does not load

the clips.

The recorded model spend was $23.62:

$0.96 for twelve 4K GPT Image 2 generations

$0.03 for one 1K image-to-image smoke test

$22.63 for six eight-second Seedance dives and five five-second connectors

Those numbers are the sum of the JSON receipts saved beside the generated files. They describe model spend only. They do not price the concept, prompts, continuity system, frontend work, review, or rejected directions.

A six-scene visual world with one geography, five recurring people, and one copper path carrying the story from intake constraint to operating rhythm

An eleven-clip, 73.46-second Seedance 2.0 film with frame-locked transitions

A live scroll-scrubbed site that works in both scroll directions

Responsive focal crops and a still-only reduced-motion path

A recorded model bill of $23.62

Northline is a production study, so the result is the inspectable film, the live interaction, and the receipt. It has no client, user, revenue, or conversion metric.

The model bill was $23.62. Keeping one world coherent across eleven clips took the continuity contract, boundary-frame connectors, and the scroll engine. That is the system I would reuse on a real engagement.

If your offer needs a working visual prototype people can see and explore, I would like to hear what you are building.