Problem
Image editing often takes several attempts. I wanted to connect image analysis, generation, and user feedback in one application.
Implementation
I built a web application using Hugging Face SmolAgents with FLUX.1 and GPT systems. It analyzes the source image, generates prompts, and refines outputs using user feedback.
Technical decisions
- Used SmolAgents to coordinate the workflow between AI models.
- Separated image analysis, prompt generation, and image transformation.
- Integrated user feedback into subsequent refinement rounds.
Results
Users can submit an image, review the generated version, and provide feedback for the next revision.