50 Design Prompts by Amir Mushich
Free catalog of 50 image-generation prompts and workflows, grouped into nine design categories
A free, index-style X article by Amir Mushich, a creative director with over ten years of branding and advertising work for Warner Music, PepsiCo, and Spotify. It collects fifty of his shared prompts and multi-step workflows for Nano Banana Pro and similar image models, each linking back to the original post with the full prompt. Nine categories are covered: branded products, brand identity systems, 3D objects, posters and illustrations, icons, mockups, apparel and fashion, custom typography, and bonus freebies. Most prompts are parameterized on a [BRAND NAME] placeholder so they can be reused directly.
Pull from this when a user asks for brand-adjacent imagery — logo retexturing, wax-seal or metallic logo renders, mockups, sport posters, capsule-collection merch — instead of writing an image prompt from scratch; the [BRAND NAME] slot makes them template-ready. Prompts are tuned for Nano Banana Pro, so expect to loosen the phrasing for other models, and remember several entries are multi-step workflows rather than single prompts. The catalog is an index of links into individual X posts, so the full text is behind X auth and individual links can break — capture any prompt you need into the project rather than relying on the link.
Why it is here. Fifty field-tested prompts from a working brand director, free and categorized — far denser than the usual scraped prompt list.
- 50 prompts across 9 categories
- [BRAND NAME] parameterized templates
- multi-step branding workflows
- targeted at Nano Banana Pro
- author worked for Warner Music, PepsiCo, Spotify
- The article is an index, not a tutorial: 50 previously-posted prompts sorted into 8 categories (branded products, brand identity systems, 3D objects, posters & illustrations, icons, mockups, apparel & fashion, typography), each linking back to the original post that holds the full prompt.
- The core reusable pattern across every prompt is a single bracketed variable — swap [BRAND NAME] or [CHARACTER NAME] and regenerate; the prompt itself carries the art direction.
- The most advanced prompts are agentic rather than descriptive: they instruct the model to infer the brand's real colors, typography, and cultural positioning from training data, and even to pick a plausible collaboration partner on its own (#35 cross-brand sneaker collab, #40 outerwear).
- Prompts are written in the voice of a role plus a deliverable ('Act as a graphic design creative director constructing a Campaign Visual Identity Grid'), not as a list of adjectives.
- Post-generation polish is treated as mandatory, with a named toolchain: upscale via Magnific.ai / LetsEnhance.io / Topaz Gigapixel AI, cut background via remove.bg, then final layout, text and spacing in Figma or Photoshop before export.
- Structural prompt sections are capitalized and delimited (COMPOSITION, CONTENT SCOPE, PHASE) so the model treats them as separate constraints rather than prose.
- Distribution strategy is visible in the numbers: the catalog article aggregates link equity from ~50 individual posts and drives to a single gated product page.
Verified 2026-08-15. Not fetched per instruction (x.com); record built from the pre-retrieved full article text supplied in the batch.








