Takes text prompt as a input and generates a image using runwayml/stable-diffusion-v1-5 model (Diffusion model)

This project is a Stable Diffusion-based text-to-image generator with a Streamlit frontend and an optimized backend using PyTorch and Hugging Face Diffusers.

• Generate images from text prompts using Stable Diffusion.
• Choose inference steps & guidance scale for better results.
• Supports both CPU and GPU execution (automatically selects based on hardware).
• Optimized for low VRAM GPUs (MX450, RTX 3050, etc.).
• Streamlit UI for easy interaction.

git clone https://github.com/your-repo/text-to-image-generator.git
cd text-to-image-generator

pip install -r requirements.txt

streamlit run frontend.py

python img_gen.py

1. Enter a text prompt in the Streamlit UI.
2. Adjust the number of inference steps (default: 20, lower for speed, higher for quality).
3. Adjust the guidance scale (default: 7.5, lower for diversity, higher for accuracy).
4. Click 'Generate Image' and wait for processing.
5. View & save the generated image.

• The script automatically detects GPU availability.
• To force CPU execution (if GPU is slow):

  generate_image(prompt, num_steps=15, device="cpu")

• Change model to Stable Diffusion 1.4 (lighter version):

  model_id = "CompVis/stable-diffusion-v1-4"

• Reduce num_steps to 10-15 for faster generation.

• Python 3.8+
• NVIDIA GPU with CUDA (Recommended)
• Minimum 8GB RAM (16GB Recommended for CPU mode)
• Libraries: See requirements.txt

| Issue > Solution | |-----------|-------------| | Slow generation (10+ min) > Reduce num_steps to 10-15, use CompVis/stable-diffusion-v1-4 | | GPU crashes due to VRAM > Use CPU mode (device='cpu'), enable enable_attention_slicing() | | Black images generated > Increase num_steps, try different guidance_scale values | | xformers error > Remove xformers from requirements.txt and install manually |

This project is open-source under the Appache License.