Summary
Creator Youri van Hofwegen breaks down AI video prompting into a structured five-level framework, using the Higsfield platform and Cadence 2.0 model throughout to demonstrate how prompt sophistication directly and measurably impacts output quality, consistency, and creative control.
Starting at Level 1 with simple one-sentence prompts, the video shows how Cadence 2.0 fills creative gaps unpredictably — generating technically acceptable but inconsistent shots that look completely different on successive runs. Each subsequent level adds specificity: Level 2 introduces subject, action, setting, lighting, and mood descriptors; Level 3 adds camera language including lens type (long-lens compression is highlighted as a frequently overlooked variable that fundamentally changes a shot’s feel); Level 4 introduces timestamped multi-shot sequences with explicit subject locks, cross-frame consistency rules, and negative prompts — a feature most beginners skip entirely. All prompts are published in the video description for viewers to replicate.
A recurring insight across levels is the importance of showing mood through specific physical behavior rather than naming it — writing “his jaw is set, his gloves tap together as he walks” rather than simply “tense.” While the framework is demonstrated on Cadence 2.0, the principles around subject locking, lens specification, timestamp-driven sequencing, and negative prompting are transferable to Runway, Kling, and other leading video generation platforms.
📺 Source: Youri van Hofwegen · Published August 06, 2026
🏷️ Format: Tutorial Demo
