Single image, 3-7 iterations with AI critics: Opus 5.5 output 'wildly better' than the author's own work
mtizard · x · 2026-10-21
Developer @dvsch shared a creation workflow using Opus 5.5: a single image reference, 3-7 iterations, with Astra/Fable acting as critic models. He calls the results imperfect but 'wildly better than I could ever do myself.'
It's a hands-on example of a multi-model loop — generator plus critics — to iteratively improve AI-generated output.