Text to Sketch
Sketch to Image
Sketch to Mood board
Sketch to Mood board 2

Inspiration

In fashion, ideas move fast but tools often slow creators down. We imagined a world where a designer can go from a rough sketch to a full presentation in minutes, not hours. Picture this: you are a designer preparing for Paris Fashion Week 2026, sketching a stunning silk gown while the deadline ticks closer. Coloring, styling, and visualizing take forever. That is the moment when FashionMoodBoard Agent steps in.

We wanted to create an AI-powered studio assistant that feels like a real collaborator. A system that understands creative intent, fills in details, and delivers professional-quality results instantly. The inspiration came from real designer workflows in high-pressure environments where imagination must meet deadlines.

What It Does

Upload a sketch or type a text description, and our AI brings it to life using Gemini Nano, Gemini Flash, and Veo 3.

The system adds colors that pop such as ocean blue, fiery red, and forest green with cinematic detail and fashion realism.

It automatically generates a mood board filled with matching accessories, textures, and runway looks.

Want to see it in motion? Veo 3 simulates a runway model wearing your design under bright lights.

The result is a complete fashion presentation with outfits, accessories, and atmosphere, ready to share, pitch, or post.

How We Built It

We developed a multi-agent system using Google ADK and Gemini APIs to coordinate multiple creative processes.

Supervisor Agent plans and manages the workflow between other agents.

Text-to-Sketch Agent produces pencil-style outlines from prompts or uploaded references.

Sketch-to-Digital Agent refines sketches into high-resolution digital illustrations using Gemini Flash.

Moodboard Agent compiles the final visuals, adds accessories, and arranges the composition.

Runway Agent (Veo 3) animates the finished design into a short runway video clip.

The user interface, built directly on Google Cloud ADK’s front-end framework, supports multi-agent chat, file uploads, and real-time voice-to-text input through Google Cloud Speech-to-Text. The entire system runs in a reproducible Docker environment within ADK Studio for smooth testing, debugging, and deployment across Google Cloud infrastructure.

Challenges We Ran Into

Coordinating multiple AI agents required precise context sharing. Each step needed clear inputs and outputs so creativity stayed structured. Another challenge was balancing freedom and control to ensure every generation looked fresh while maintaining visual consistency. Latency was also a concern, especially with video rendering from Veo 3, so we implemented streaming updates to provide live feedback and partial previews while models completed their outputs.

Accomplishments That We’re Proud Of

We created a full pipeline that turns a plain sketch into a cinematic fashion showcase. The platform feels responsive and intelligent, able to interpret creative briefs and deliver visuals that match human design instincts. We are proud that we achieved a completely cloud-based setup using Google Cloud ADK, which allows scalability and easy collaboration. Designers can now drag in a sketch, describe their idea, and receive a complete mood board and runway preview in a single session.

What We Learned

We learned that structured AI collaboration works best when every agent has a clear role. Multi-agent systems require small, modular components that communicate efficiently. We also learned how much real-time feedback improves usability; progress updates and partial previews make users feel involved in the process. Working with Veo 3 and Gemini Flash gave us a deeper understanding of motion rendering and visual coherence. Voice input integration proved that natural interaction can make creative ideation faster and more intuitive.