Krea-2 Turbo in ComfyUI: composites and character tutorial

Nerdy Rodent demonstrates Krea-2 Turbo workflows in ComfyUI for combining reference images and generating character prompts. The tutorial builds on an earlier workflow rather than covering installation.

Two approaches drive the compositing examples: image compositing with a mask value to control the mix, and latent blending before sampling. The speaker combines sketches, cartoons and space scenes while keeping the main prompt unchanged in several examples. For latent blending, both images must have identical dimensions. The demonstrated workflow stretches them to 1280 by 720 and uses a denoise value of 0.65 so the blended latent influences the result.

The tutorial also tests NegPip and a rebalance guider for negative prompting. The speaker reports changes in the examples but cautions that neither method produces perfect negative prompting. Higher text scale values in the advanced enhancer tend to produce more text in his tests, without necessarily improving its coherence.

The final section uses Identity Forge to assemble random character prompts or select an archetype such as Arcane Wizard. Despite its placement in an AI prompting group, the speaker says the node pack itself does not use AI. Concatenating its output with a user prompt adds control over image style and setting.