Qwen Image HPC is a Gradio web application for text-to-image generation using the Qwen/Qwen-Image diffusion model. It features flexible image generation controls, streamlined UI, easy batching, example prompts, and robust backend support for high-performance computing environments with GPU acceleration.

• Prompt-Based Image Generation: Enter descriptive prompts to create high-quality images.
• Negative Prompts: Optionally guide the model to avoid specific elements in your images.
• Aspect Ratio Selection: Choose from common aspect ratios (1:1, 16:9, 9:16, 4:3, 3:4) with automatic size updates.
• Seed Control: Fix or randomize seed for reproducible or varied outputs.
• Advanced Controls: Adjust image size, guidance scale, inference steps, and batch size (number of images).
• Output Management: Download results individually or as a ZIP archive.
• Examples: Predefined sample prompts for quick testing.
• Local File Saving: Each image is automatically saved with a unique name for reference.
• Performance Tracking: Displays seed used and generation time.
• UI Customization: Simple, compact theme ideal for production or demonstration use.

• Backend: Uses Diffusers' DiffusionPipeline to load "Qwen/Qwen-Image" with bfloat16 on CUDA (or CPU fallback).
• Frontend: Built with Gradio Blocks, features prompt input, aspect ratio dropdown, advanced settings accordions, and result galleries.
• Helper Functions: Manage random seed, image saving, and ZIP packaging.

1. Clone the repo:

   git clone https://github.com/PRITHIVSAKTHIUR/Qwen-Image-Diffusion.git
   cd Qwen-Image-Diffusion
   

2. Install dependencies:

   pip install git+https://github.com/huggingface/transformers.git
   pip install git+https://github.com/huggingface/diffusers.git
   pip install git+https://github.com/huggingface/accelerate.git
   pip install git+https://github.com/huggingface/peft.git
   pip install huggingface-hub sentencepiece safetensors gradio torch pillow numpy

python app.py

The app will launch a Gradio web interface. Enter your prompt and generate images!

• Realistic Still Life: "Realistic still life photography style: A single, fresh apple, resting on a clean, soft-textured surface..."
• Chinese Painting: "一幅精致细腻的工笔画，画面中心是一株蓬勃生长的红色牡丹..."
• Classroom Scene: "A young girl wearing school uniform stands in a classroom, writing on a chalkboard..."
• Hand-drawn Water Cycle: "手绘风格的水循环示意图，整体画面呈现出一幅生动形象的水循环过程图解..."
• Capybara Mascot: "A capybara wearing a suit, holding a sign, that reads Hello World"

• Aspect Ratio Quick Set: Easily adjust width and height sliders to match standard ratios.
• Batch Generation: Create 1–5 images per prompt and package output as a ZIP.
• UI Tweaks: Modify the CSS and theme settings in app.py as needed.

https://github.com/PRITHIVSAKTHIUR/Qwen-Image-Diffusion

Note: Requires a machine with a supported NVIDIA GPU for best performance. For research and educational use.