Designer mixes multiple AIs

A designer’s post about mixing Midjourney, Gemini and Krea shows how some brand work is shifting from one generator to a stack of tools used in sequence. (x.com) The workflow described in the post uses Midjourney, Gemini and Krea to test multiple visual identity directions quickly, then relies on the designer to choose what survives. Krea says its Realtime tool updates output “immediately” as a user draws, types or moves elements on a canvas. (x.com) (docs.krea.ai) That stack reflects how the products are built. Midjourney describes itself as a “community-funded research lab” focused on AI image models, while Google says Gemini Image supports conversational image creation and editing with detailed prompts and uploaded references. (midjourney.com) (deepmind.google) In plain terms, the tools do different jobs. Midjourney is often used for stylized visual options, Gemini can iterate with text and reference images across multiple turns, and Krea is built for live exploration on a canvas instead of one-shot prompting. (midjourney.com) (deepmind.google) (docs.krea.ai) That matters for visual identity work because branding usually starts with volume, not a single finished mark. Google’s image tools allow up to 14 reference images in Gemini workflows, and Krea’s Realtime docs say rough strokes and loose color blocks can steer generations before a designer refines the result. (ai.google.dev) (docs.krea.ai) The post also draws a line between generation and authorship. Google says its Imagen models add SynthID watermarks to generated images, while the designer’s point was that taste, selection and editing still sit with the human making the final call. (ai.google.dev) (x.com) Designers have used mood boards, sketches and reference pulls for decades; the change is speed. Krea says its canvas has “no queue, no waiting, and no render button,” which turns ideation into a live back-and-forth instead of a slower export cycle. (docs.krea.ai) The tradeoff is that more options do not remove the need for judgment. Midjourney, Gemini and Krea can all generate images, but none of their official materials claim to decide which concept fits a client, a market or a brand system. (midjourney.com) (deepmind.google) (krea.ai) So the post lands less as a claim that one model won, and more as a snapshot of a design process built around comparison. The software can flood the canvas with directions; the designer still decides which one becomes an identity. (x.com)