Crevia AI

Real-Time Brand-Aware AI Design Assistant for Adobe Express

Comment 1

Crevia AI – Overview

Inspiration

Designing for social media is fast-paced and high-stakes, but most creators and marketers struggle to keep every post on-brand, accessible, and engaging. Manual review is slow and subjective, and even experienced designers miss details. We wanted to build a tool that gives instant, actionable, and brand-aware feedback, right inside Adobe Express, so anyone can create better content, faster.

What it does

Crevia AI is a real-time design assistant for Adobe Express. It analyzes your canvas, scores your design for clarity, consistency, and engagement, and gives you AI-powered, brand-aligned suggestions. With one click, you can apply fixes (like font, color, or alignment changes) directly to your design. The connected webapp lets you manage your brand profile, see report history, and export results.

How we built it

  • Adobe Express Addon: React-based UI panel, using the Adobe Sandbox API to read and (mock) update canvas elements.
  • AI Engine: Canvas and brand data are formatted into a structured prompt for OpenAI (GPT-3.5). The AI returns a JSON report with scores and actionable suggestions. (planned)
  • Fix Engine: An executor applies suggested changes to the canvas in the sandboxed environment. (mocked for now)
  • Webapp: Built in React.js, for storing brands, reports, and users. The dashboard shows report history, brand profiles, and score trends.
  • Export: JSON/PDF/PNG export is in progress.

Challenges we ran into

  • Adobe Sandbox API limitations: Not all element types are editable yet, so some fixes are mocked.
  • Prompt engineering: Getting reliable, structured output from the AI required careful prompt design and fallback handling for malformed responses.
  • Brand data integration: Ensuring the AI always respects user brand guidelines, even with incomplete or inconsistent data.
  • Export formatting: Generating branded, readable PDF/PNG exports that look good across devices.

Accomplishments that we're proud of

  • End-to-end pipeline: From canvas parsing to AI suggestions to (mock) fixes, all in one flow.
  • Real brand data: Brand fonts, colors, and logo are injected into the AI prompt for truly personalized feedback.
  • Interactive UI: Users can see suggestions, apply fixes, and export reports—all in a polished, production-ready interface.
  • Webapp integration: Brand management and report history are live and connected to the addon.

What we learned

  • AI output is only as good as the prompt and data you provide. Iterating on prompt structure and fallback logic was key.
  • Designers want control: Users appreciate being able to preview and selectively apply fixes, not just "auto-fix everything."
  • Brand consistency is a moving target: Every team has different standards, so flexibility in brand settings is crucial.
  • APIs evolve: Building with the Adobe Express SDK means planning for future capabilities and workarounds.

What's next for Crevia AI

  • Shipping in 4-6 weeks post-hackathon!
  • and many more! follow us on getcrevia.com

Built With

Share this project:

Updates