NovaVision

Generate art based on how you're feeling. Type a sentence, and it detects the emotion, builds a visual prompt, and creates an image using FLUX.1.

Demo

How it works

Text Input → Emotion Detection → Prompt Engineering → FLUX.1 Generation → Artwork
 "I feel     (HuggingFace       (emotion → visual    (diffusion model      (unique
  peaceful    Transformers)       style mapping)       via HF Inference)     image)
  today"

Supported emotions: joy, sadness, anger, fear, surprise, disgust, neutral — each mapped to different visual styles, color palettes, and artistic techniques.

Example

Input:  "The sunset made me feel completely at peace with the world"
→ Emotion: Joy (confidence: 0.94)
→ Generated prompt: warm golden light, serene landscape, soft brushstrokes...
→ Output: [serene landscape in warm golden tones with soft brushstrokes]

Quick start

git clone https://github.com/urme-b/NovaVision.git
cd NovaVision
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt

# Set up HuggingFace credentials
cp .env.example .env
# Edit .env with your HuggingFace API token

python server.py
# → http://localhost:8000

Requires: Python 3.10+, HuggingFace account with API token

API endpoints

Method	Endpoint	Description
POST	/api/analyze	Detect emotion from input text
POST	/api/generate	Full pipeline — detect emotion + generate artwork

Example request

curl -X POST http://localhost:8000/api/analyze \
  -H "Content-Type: application/json" \
  -d '{"text": "I feel so grateful for everything today"}'

{
  "success": true,
  "primary_emotion": "joy",
  "confidence": 92.0,
  "valence": 0.8,
  "arousal": 0.7
}

Repo structure

NovaVision/
├── src/
│   ├── models/            # Data schemas and type definitions
│   └── services/          # Emotion analysis and image generation
├── config/                # Application settings (Pydantic)
├── tests/                 # Unit and integration tests (pytest)
├── server.py              # Flask API server
├── app.py                 # Gradio web interface
├── index.html             # Single-page web app
└── requirements.txt

Tech Stack

Python · Flask · HuggingFace Transformers · FLUX.1 · Pydantic · Pytest · Docker

Testing

pytest tests/ -v

Roadmap

• User authentication and gallery history
• HuggingFace Spaces deployment for a public demo
• Batch processing for multiple text inputs
• Style customization (choose artistic influences)

License

MIT