Fredrik steered the creative work and reviewed the results. ChatGPT coordinated the tool-assisted production: prepare shot instructions, request model runs through Fal, review takes, assemble the edit and check the export.
Give each model a bounded job
- Kling 4.0
- Image-to-video motion for the atelier, camera work and other cutaways.
- Kling Avatar v2 Pro
- Performance takes driven by short audio guides from the song.
- Sync Lipsync
- Version 2 Pro and version 3 were run during refinement of the opening performance.
The cooperation moves through concrete handoffs: a shot brief becomes a generated take; a selected take and its audio guide become a lip-sync task; the reviewed clips become inputs to the edit.
Keep the music continuous
The film uses the song’s 03:16–04:36 passage as one continuous 80-second audio bed. Visual cuts follow the music while performance takes use the corresponding phrases as guides. The finished presentation runs at 24 frames per second in 1280 × 720.