ProductVision Studio

• 

  "AI-Powered Product Photography Studio - From Concept to Professional Images in Minutes"

Inspiration

We were inspired by the challenge of creating professional product photography for e-commerce and marketing. Traditional product photography is expensive, time-consuming, and requires specialized equipment. We wanted to democratize professional product imagery using AI.

What it does

ProductVision Studio generates professional product photography using multi-agent AI orchestration. It offers 2 generation modes: Professional (with reference images and presets) and Simple (free-form), and demo generation is present to see what kind of product images it can generate, Admin Demo (for judges) and User Demo(36 images, 3 products, 12 variation). Users get studio-quality 1024x1024 images with different camera angles and lighting setups.

How we built it

We built a multi-agent system using Gemma 3:4b for generating Json schema/prompt to give to FIBO image model for generation. We have used Ollama to run Gemma3:4B model locally. The application takes user inputs and gives AI the prompt and it creates JSON schemas with camera angles, lighting, and composition details. The generation agent uses ComfyUI with BRIA nodes to create images. We implemented sequential VRAM management to handle GPU VRAM limitations and created a real-time progress tracking system so that who wants to use FIBO model locally rather than API can also use it smoothly.

Challenges we ran into

VRAM Management: 6GB GPU limit required sequential processing (Ollama(Gemma3:4B → BRIA AI(FIBO)) Schema Integration: Converting AI-generated schemas to ComfyUI workflows Real-time Tracking: Accurate progress monitoring across multiple generation phases Reference Images: Building pipeline from base64 → temp files → ComfyUI LoadImage nodes Performance: Optimizing UI animations for low-end devices while maintaining quality

Accomplishments that we're proud of

Multi-Agent Orchestration: Seamless coordination between planning and generation agents Professional Controls: Reference image support, lighting presets, and schema preview system Award-Winning UI/UX: Premium SaaS design with glassmorphism and performance optimization Production Ready: Complete application with 4 modes, history, and comprehensive documentation Real Innovation: JSON-native control and sequential VRAM management

What we learned

Multi-agent AI systems require careful orchestration and error handling VRAM management is crucial for consumer GPU deployment Real-time progress tracking significantly improves user experience Schema preview systems give users control over AI generation Performance optimization doesn't require sacrificing visual quality

What's next for ProductVision Studio

Cloud Deployment: Scale with multiple ComfyUI instances and load balancing Custom Products: User-uploaded product images with AI enhancement Video Generation: Animated product showcases and 360° views API Access: Developer API for e-commerce platform integration Style Transfer: Brand-specific styling and consistency controls

Built With

• bria-ai
• comfyui
• framer-motion
• lucide-react
• next.js
• ollama-(gemma-3:4b)
• shadcn/ui
• tailwind-css
• typescript