Descriptions:
Youri van Hofwegen walks through the complete production workflow he used to create a 16-minute AI short film, organized around a single core principle: never start with video generation. Most beginners, he argues, give AI video tools too much creative freedom — characters drift between scenes, details go missing, and the model fills in gaps inconsistently. The fix is to build comprehensive reference images for every visual element first, so the video generator has nothing left to invent.
The workflow uses Higgs Field Cinema Studio, a platform trained exclusively on real film footage. Van Hofwegen builds two characters (Ryder and Vance) using different methods: Ryder via a stock-image reference for exact look matching, Vance using the platform’s AI cast feature which accepts genre, budget scale, and visual trait selections to generate a character from scratch. Locations follow the same discipline — a desert canyon bridge is generated cleanly, then a second destroyed version is produced by feeding the original as a reference image and prompting only for damage, keeping structure and lighting identical. Armored convoy vehicles are generated using Soul Cinema 2K, a separate model optimized for hyperrealistic shots.
With all references locked, the video director panel handles camera movement, speed ramps, clip duration, and resolution. Van Hofwegen argues that front-loading roughly 80% of creative decisions into the reference phase makes video generation significantly more predictable and controllable — a transferable principle for anyone building AI-assisted narrative content.
📺 Source: Youri van Hofwegen · Published June 01, 2026
🏷️ Format: Tutorial Demo







