Krita AI Diffusion: Z-Image Turbo and Flux 2 Klein setup

This tutorial covers Krita's AI Diffusion plugin, focusing on Z-Image Turbo and Flux 2 Klein in version 1.47, with a closing update on version 1.48. The setup includes installing the plugin ZIP, enabling its dock panel, and choosing a local ComfyUI server or a paid cloud connection. The speaker recommends installing workflows through the main tab after encountering missing modules with individual component installation.

For local AI users affected by the speaker's Flux 1 loading problem, the walkthrough explains a workaround: download alternative FP8 checkpoints, place them in the diffusion_models directory, and duplicate the Flux styles to select those files. The editing style also needs to point to the copied Flux Kontext configuration.

The composition exercise combines masked generation with background replacement and instruction-based edits. The speaker finds Z-Image Turbo consistent but limited in variety, with unreliable text toward the ends of sentences. They describe Flux 2 Klein as faster and lighter than Flux 1 and more tolerant of repeated edits than Kontext, while still preferring Kontext for some precise instructions.

The final examples cover fill, expand, and color match in version 1.48. The speaker reports less haze around generated edges and advises masking around text that must survive an edit. The source description notes that most generation footage runs at 8x speed or faster.