Text to Image Generator

A Python-based text-to-image generation tool leveraging stable diffusion models. This application provides enhanced control over image generation through optimized parameters, improved prompting, and an efficient interface.

Features include weighted negative prompts, dynamic thresholding, optimized VAE processing, and memory-efficient operations. Available both as a local application and Google Colab notebook for flexible deployment options.

Key Improvements

Enhanced Negative Prompts

• Weighted emphasis (1.4x, 1.5x) on critical issues
• Comprehensive categories with better organization
• Stronger focus on preventing duplications and deformities

Optimized Parameters

• Increased default steps to 50 for better quality
• Adjusted guidance range (7-30, default: 15)
• Added dynamic thresholding
• Enhanced scheduler configuration

Improved Prompt Weighting

• Main subject emphasis: 1.4x
• Secondary elements: 1.2x
• Better distribution of emphasis

Technical Optimizations

• Sequential CPU offload
• Automatic attention slicing
• Enhanced VAE optimization
• Improved memory management

UI Enhancements

• Extended quality steps range (30-150)
• Better step increments
• Added information tooltips
• Improved layout and descriptions

Expected Results

• Higher quality images with fewer artifacts
• No duplications or deformed bodies
• More accurate prompt following
• Better overall consistency

Setup Guide

Google Colab Setup

1. Open Text-to-img.ipynb in Google Colab
2. Select 'Runtime' > 'Change runtime type' > Choose 'T4 GPU'
3. Run all cells in sequence

Hugging Face Token Setup

Google Colab

1. Obtain your Hugging Face token from Hugging Face.
2. Add the token to the Colab notebook by running the following code in a cell:

   from huggingface_hub import login
   login("your_huggingface_token_here")

Local Deployment

1. Obtain your Hugging Face token from Hugging Face.
2. Create a file named token inside the huggingface directory:

   echo "your_huggingface_token_here" > huggingface/token

Alternatively, you can securely add your Hugging Face token directly in Google Colab by setting it as a secret. Use the following code snippet in a cell:

HF_TOKEN=YOUR_HUGGINGFACE_TOKEN

Local Installation

1. Clone Repository

git clone https://github.com/SauRavRwT/text-to-img.git
cd text-to-img

2. Create Virtual Environment

Windows:

python -m venv venv
venv\scripts\activate

Linux/Mac:

python -m venv venv
source venv/bin/activate

3. Install Dependencies

pip install --upgrade pip setuptools wheel
pip install -r requirements.txt

4. Launch Application

python app.py