Summary
Youri van Hofwegen walks through a complete hands-on tutorial of Pixels, an AI tool built specifically for YouTube thumbnail creation. Unlike general-purpose image generators, Pixels is trained on thumbnail psychology — what makes viewers click — and scores every output across five dimensions: virality, clarity, idea, curiosity, and emotion. Van Hofwegen demonstrates that meaningful thumbnails can be generated from minimal text prompts, with results scoring between 94 and 98 out of 100 across most use cases.
The video covers four distinct workflows: generating thumbnails from scratch using natural-language prompts; using Pixels’ one-click fix feature, which rebuilds underperforming images based on its own scoring analysis (demonstrated by taking a horror-concept thumbnail from a score of 76 to 94 without rewriting the prompt); cloning the proven layout of an existing high-performing thumbnail while swapping in new subjects (a Conor McGregor example achieved the session’s top score of 98); and using Pixels’ persona feature, which trains on uploaded photos to insert the creator’s face — or an AI influencer avatar — into any generated thumbnail.
Van Hofwegen also addresses text rendering, a well-known weakness of most image generators, showing that Pixels handles spelled-out copy correctly and places it compositionally. The tutorial is aimed at YouTube creators who want to reduce design time without hiring a graphic designer, and the walkthrough is practical enough that viewers can replicate the workflow immediately using the prompts provided in the video description.
📺 Source: Youri van Hofwegen · Published August 23, 2026
🏷️ Format: Tutorial Demo
