🧠 Prompt Engine

A powerful AI-based system that converts text, image, and audio inputs into high-quality, structured prompts for generative AI models like Stable Diffusion, Midjourney, and DALL·E.

---

🚀 Features

• ✍️ Text → Prompt Refines and extends simple prompts into detailed, high-quality prompts.

• 🖼️ Image + Text → Prompt Understands an image and user intent to generate a descriptive prompt.

• 🎧 Audio → Prompt Converts speech into text and then generates a refined prompt.

• 🧠 Multimodal AI (Janus-Pro-1B) Uses a vision-language model for intelligent prompt generation.

• 🎨 Gradio UI Interactive web interface for easy usage.

---

🧩 Architecture

Input (Text / Image / Audio)
        ↓
Preprocessing Layer
  (Whisper for audio)
        ↓
Instruction Builder (Prompt Engineering)
        ↓
Janus-Pro-1B Model
        ↓
Post-processing (clean output)
        ↓
Final AI Prompt

---

🛠️ Tech Stack

• Python
• HuggingFace Transformers
• DeepSeek Janus-Pro-1B
• OpenAI Whisper (Speech-to-Text)
• Gradio (UI)
• PyTorch

---

📦 Installation

1. Clone the repository

git clone https://github.com/your-username/prompt-generator.git
cd prompt-generator

---

2. Install dependencies

pip install -r requirements.txt

---

3. Run the application

python app.py

---

🧪 Usage

1. Open the Gradio UI in your browser

2. Select input type:

   • Text
   • Image + Text
   • Audio

3. Provide input

4. Click Generate Prompt 🚀

5. Get your refined AI prompt

---

🧠 Example

Input:

boy in forest

Output:

A cinematic scene of a young boy standing in a dense forest, soft sunlight filtering through tall trees, atmospheric fog, ultra-detailed, 4k, depth of field, masterpiece

---

📁 Project Structure

project/
│
├── app.py                
├── requirements.txt
└── README.md

---

⚙️ Core Functions

• text_to_prompt()
• image_text_to_prompt()
• audio_to_prompt()
• generate_universal_prompt()

---

⚠️ Limitations

• Requires GPU for best performance
• Video input not supported (yet)
• Output quality depends on prompt instruction

---

🔮 Future Improvements

• 🎥 Video input support
• 🎨 Style selection (anime, cinematic, realistic)
• 📊 Prompt scoring system
• ☁️ Deployment on HuggingFace Spaces

---

🤝 Contributing

Pull requests are welcome! For major changes, please open an issue first.

---

📜 License

This project is open-source under the MIT License.

---

👨‍💻 Author

Anshu Singh

---

⭐ If you like this project

Give it a ⭐ on GitHub!