From Reference Image to Brand System: Amir Mušić's AI Agent Workflow

X postAmirMušićTutorial

Designer Amir Mušić shares a workflow and the open-source 'Lovart Brand System Skill', enabling the Lovart Agent to build a full brand identity from a single visual reference. Rather than copying aesthetics, the agent deconstructs underlying visual logic and applies it across guidelines, campaign assets, and packaging concepts.

Key points

  1. Deconstruction over copying: Using a reference ('Pickle Potion'), the workflow extracts structural rules—such as split-screen composition, bold typography, and real ingredient photography—and translates them to a new brand concept ('Juicee').
  2. Curating the Brand Anchor: The designer rejected early agent drafts that mixed conflicting styles, refining them down to an approved 'Brand Kit' covering logo, typography, and colour rules.
  3. The 'Brand Lock' methodology: To counter visual drift when introducing new scene references, a strict control hierarchy dictates which reference governs composition, colour, or product identity.
  4. Tooling and licensing: Built with Lovart Agent and refined using features like Text Edit and Touch Edit (with Seedance 2.5 used for motion), the Brand System Skill is published on GitHub under CC BY 4.0.

Why it matters

Generative AI often struggles with cross-asset consistency. Mušić's methodology highlights how systematic constraints and human creative direction allow autonomous design agents to produce coherent, multi-SKU brand worlds rather than disjointed one-off visuals.

Analysis

Mušić clarifies that an AI agent requires creative direction: the primary task was not generating more outputs, but rejecting inconsistencies and enforcing hierarchy. He also clarifies that generated packaging views serve as conceptual mockups rather than production-ready dielines or vector prepress files.

Original post

AmirMušić

@AmirMushich · 3 September 2026

the only Design_Skill you need to make your agent produce studio quality branding GitHub link - in the article below: t.co/3m2XvAMxAn t.co/RwE3y1VKmv

657 likes15 replies1.3k bookmarks64.7k views

View on X
More posts in the thread (4)
  1. @0xTreff appreciate you Treff
  2. @ItsNazar17 Yessir! The Skill itself is licensed under CC BY 4.0, so you can use it commercially, adapt it, and implement it in your app. Just credit me and the original repository, link the license, and clearly note any changes you make Any third-party assets or services remain subject
  3. @Awesome_O_AI fax
  4. @andriy_uiux absolutely brother, I'm on a mission to leave everything I learned about design to the people <3
Selected replies (5)
  • @ItsNazar17 @AmirMushich Whenever I see a recommendation with "Amir" and "Design" words in it, I immediately open the article/post to read it. This skill looks really valuable, I'm definitely going to try it out. One question, since it's publicly available, can I use it commercially, like implementing
  • @0xTreff @AmirMushich The fact that it ships with a benchmark and changelog makes it feel like an actual product Not just a prompt file
  • @Awesome_O_AI @AmirMushich This is the missing bridge between generic generation and repeatable studio quality
  • @andriy_uiux @AmirMushich Thanks man, appreciate you share it open source
  • @DigitalToks @AmirMushich @iksly2

Mentioned links

Links and tools from this entry

Summary generated by AI (Gemini 3.8 Flash, 27 September 2026). It may contain errors – the original source is authoritative.

Related entries

Similar in content, found by the AI search.

  • X post:klöss

    From Reference Image to Brand System: AI-Driven Branding Workflow

    klöss highlights an open branding workflow published by designer Amir Mušić, built on over a decade of industry experience. Mušić details how an AI agent (Lovart Agent) can turn a single reference image into a complete brand system, including guidelines, conceptual packaging, and multi-asset campaign visuals maintained via a 'Brand Lock' framework.

    471bookmarks32.7kviews

    AI & LLMs· Resource

  • Link:amirmushich.link

    Lovart: AI Design Agent Platform Featuring Layer Editing and Multi-Model Routing

    Lovart is a web-based creative platform functioning as an AI design agent to generate on-brand assets from briefs and brand guides. The application combines prompt-driven generation with granular post-processing tools, such as editable typography layers, object isolation, and unified access to leading image and video AI models.

    AI & LLMs· Tool

  • Link:Resend

    Resend Publishes Design System Guidelines and Agent Skill

    Resend has documented its design and brand guidelines in a machine-readable format, packaging the specifications as an installable Agent Skill alongside a web reference. The guide defines typography, colours, logo constraints, and core design principles.

    Web development· Resource

  • Article:Vercel

    How Vercel Teaches Coding Agents Product Design

    Vercel details an internal framework called product-design that equips coding agents with the product context and judgment needed to build coherent interfaces.

    AI & LLMs· Resource

  • Link:Emil Kowalski

    Emil Kowalski's Design Engineering Skill for AI Agents

    Emil Kowalski has published an installable agent skill based on his design engineering philosophy. It aims to guide AI coding assistants in building polished, detail-oriented user interfaces.

    AI & LLMs· Resource

Back to the knowledge base

Let's talk about your project

Get in touch

Locations

  • Mattersburg
    Johann Nepomuk Bergerstraße 7/2/14
    7210 Mattersburg, Austria
  • Vienna
    Ungargasse 64-66/3/404
    1030 Wien, Austria

Parts of this content were created with the assistance of AI.