Visionary-AI — AI Image Generator

Visionary-AI is a small, flexible toolkit for generating images from text prompts using modern image-generation models. It provides a CLI and a minimal web UI (if included in the repo) to quickly turn prompts into high-quality images, plus sensible defaults and configuration for local or cloud inference.

Note: This README is a general, practical guide. Adjust environment variables, commands, or model names to match the implementation in the repository if necessary.

Features

• Generate images from text prompts
• Support for configurable model backends (local / cloud API)
• Batch generation and prompt templates
• Simple CLI and optional web UI
• Output management: set resolution, number of steps, seeds, and output folder
• Configurable via environment variables or a config file

Table of Contents

• Installation
• Quickstart
• Usage
  • CLI
  • Web UI
  • Python API
• Configuration
• Examples
• Developing
• Contributing
• License
• Acknowledgements

Installation

Prerequisites:

• Python 3.8+ (or the version used in this repo)
• pip
• (Optional) A GPU + CUDA for local model inference

From the repository:

1. Clone the repo git clone https://github.com/mohas8/Visionary-AI-AI-Image-Generator.git cd Visionary-AI-AI-Image-Generator

2. Create and activate a virtual environment (recommended) python -m venv .venv source .venv/bin/activate # macOS/Linux .venv\Scripts\activate # Windows

3. Install dependencies pip install -r requirements.txt

If the project provides a Dockerfile: docker build -t visionary-ai . docker run --rm -e MODEL_API_KEY=$MODEL_API_KEY -p 7860:7860 visionary-ai

Quickstart

Basic CLI usage (adjust command to match actual CLI script in the repo):

Generate a single image from a prompt: python -m visionary main --prompt "A photorealistic red fox sitting in a snowy forest at sunrise" --out outputs/

Generate 4 images at 512x512: python -m visionary main --prompt "Surreal painting of a city made of glass" --n_samples 4 --width 512 --height 512 --out outputs/

If the project uses an external API (OpenAI / Replicate / Stability / other), set the API key first: export MODEL_API_KEY="your_api_key_here"

Windows (PowerShell)

$env:MODEL_API_KEY="your_api_key_here"

Usage

CLI

• The repository contains a CLI entrypoint (e.g., visionary/main.py or scripts/generate.py). Run it with --help to list supported flags: python -m visionary --help

Common flags:

• --prompt : text prompt to generate from
• --out / --output-dir : directory to save generated images
• --width / --height : output image size
• --n_samples / --count : number of images to generate
• --seed : random seed for reproducibility
• --model : choose the backend/model name (if supported)

Web UI

• If a web UI is included, start it with: python -m visionary.web --port 7860
• Open your browser at http://localhost:7860

Python API

• If the repo exposes a library-like interface, you can generate images from Python: from visionary import Generator

  gen = Generator(api_key=os.environ.get("MODEL_API_KEY"), model="stable-diffusion") images = gen.generate("A fantasy castle on a floating island", width=512, height=512)

Check the repository's modules for exact class/function names.

Configuration

You can configure the tool via environment variables or a config file (config.yml or .env). Typical settings:

• MODEL_API_KEY — API key for cloud model backends
• MODEL_BACKEND — e.g., local, openai, replicate, stability
• MODEL_NAME — model identifier (e.g., "stable-diffusion-v1-4")
• OUTPUT_DIR — default directory for outputs
• DEVICE — cpu, cuda:0, etc.
• SAMPLES_PER_PROMPT — default number of images to generate

Example .env: MODEL_API_KEY=sk-xxxxxxxxxxxx MODEL_BACKEND=replicate MODEL_NAME=stability-ai/stable-diffusion OUTPUT_DIR=outputs DEVICE=cuda:0

Examples

1. Single-prompt generation: python -m visionary main --prompt "A cyberpunk skyline at night, neon reflections" --n_samples 1

2. Batch prompts from file: python -m visionary main --prompts-file prompts.txt --n_samples 2 --out outputs/

3. Reproducible output: python -m visionary main --prompt "Minimalist poster of a whale" --seed 42 --n_samples 1

Best practices

• Start with lower resolutions for fast iteration (256–512).
• Use seeds to reproduce results.
• Use prompt templates and negative prompts (if supported) to refine outputs.
• If using cloud APIs, watch your usage/costs and set rate limits if supported.

Developing

• Run tests (if included): pytest

• Linting and formatting: black . flake8

• Add new models or backends by implementing an adapter class that follows the repository's model interface (look for model/ or backends/ directory).

Troubleshooting

• Out of memory on GPU: lower width/height or batch size; switch to CPU if necessary.
• Slow generation: use a smaller model, lower sampling steps, enable caching.
• Authentication errors: confirm MODEL_API_KEY is set and valid for the chosen backend.

Contributing

Contributions are welcome. Please:

1. Open an issue describing the change or bug.
2. Create a feature branch.
3. Open a pull request with tests and documentation updates.

Follow repository coding style and include meaningful commit messages.

License

MIT

Acknowledgements

• Built with gratitude to the open-source model & community ecosystem.
• Mention any third-party libraries or model providers used in the repo (Hugging Face, Stability, OpenAI, Replicate, etc.)

Contact

For questions or help, open an issue or contact the maintainer: @mohas8