Pixray
** - A text-to-image AI generator that offers customization and unique art styles, perfect for artists and creators.
Last verified:
What is Pixray?
Pixray is a state-of-the-art text-to-image neural generation engine that transforms textual descriptions into visual artwork. It supports multiple generation technologies including VQGAN, Pixeldraw, Clipdraw, and Diffusion, allowing users to create everything from pixel art to photorealistic images. The system combines perception engines, CLIP-guided GAN imagery, and latent space navigation techniques to iteratively optimize images against classifiers.
Key features include multiple drawer types (vqgan for GAN-images, pixel for pixel art, clipdraw/line_sketch for stroke-based drawings, and diffusion for higher quality), customizable prompts with weighted importance, aspect ratio options (square, portrait, widescreen), quality presets (draft, normal, etc.), and iteration control. It supports multiple simultaneous prompts separated by pipe characters, with the ability to specify weight and stop conditions for each prompt.
Pixray is designed for developers, artists, designers, content creators, game artists, and retro aesthetic enthusiasts who want technical control over image generation. It works as a Python library and command line utility, runs in Google Colab notebooks, and integrates with Docker and Replicate. The tool is particularly valued for its genuine pixel art capabilities and customization options beyond what consumer-grade generators expose.
Pixray pricing
Pricing model: Freemium
Pixray is open-source and free to use. It offers a free tier through Replicate with initial free usage, but after a certain period a credit card is required and charges are incurred per second for predictions made at $0.0002 per second. Cost varies according to the hardware used for running different models (typically Nvidia T4 GPU). The open-source version can be self-hosted for free using Python, command line, Docker, or Google Colab.
Pixray pros
- Open-source with full code access and customization
- Multiple generation technologies (VQGAN, Pixeldraw, Clipdraw, Diffusion)
- Genuine pixel art capability, not just filtered approximation
- Supports both Python library and command line usage
- Works in Google Colab for cloud-based generation
- Multiple drawer types for different artistic styles
- Weighted prompts for fine-tuned control over output
- Aspect ratio options (square, portrait, widescreen)
- Quality presets simplify configuration
- Docker and Replicate integration for developers
- Speech-to-text functionality for hands-free prompting
- Multiple simultaneous prompts with pipe separation
- Iteration control for precise optimization tuning
- Scale parameter for increasing output size
- Active Discord community for support and discussion
Pixray cons
- Developer-oriented positioning creates setup friction for casual users
- Requires technical know-how for advanced features
- Needs 16GB RAM for optimal performance
- diffvg installation required for clipdraw drawer
- Models directory must be created manually
- Diffusion drawer needs more iterations than VQGAN
- Requires runtime restart in Colab after installation
- More configuration needed compared to polished alternatives
Frequently asked questions about Pixray
What is Pixray?
Pixray is a state-of-the-art text-to-image neural generation engine that creates images from text prompts. It supports VQGAN, Pixeldraw, Clipdraw, and Diffusion technologies, combining perception engines, CLIP-guided GAN imagery, and latent space navigation to iteratively optimize images against classifiers.
How do I install Pixray?
Clone the repository with 'git clone --recursive https://github.com/pixray/pixray', cd into pixray, run 'pip install -r requirements.txt', clone v-diffusion-pytorch and diffvg submodules, copy diffusion folder, install diffvg, and create a models directory. For Colab, run the installation cells and restart runtime after installation.
What are the different drawer types?
The drawer defines the generation technology: vqgan uses VQGAN from taming-transformers for GAN-images (often trippy or realistic), pixel uses dribnet's original pixel drawer for pixel art, clipdraw and line_sketch use diffvg for stroke-based drawings as if strokes were drawn down, and diffusion makes higher quality images but needs more iterations.
How do prompts work in Pixray?
Prompts are text that guides how the drawer generates images. Multiple prompts can be input with pipe '|' as separation. Each prompt can specify weight and stop conditions using syntax 'prompt:weight:stop'. Weight enhances importance and impact of a prompt, while stop tells the optimizer when to stop optimizing for that prompt.
What aspect ratios are supported?
Pixray supports three aspect ratios: square (size=), portrait (size=), and widescreen (size=). Specifying aspect automatically defines size for the user.
How many iterations should I use?
Typically 30 iterations show the basic layout and focus, from 100 on iterations may not change as much, and over 300 would have minimal effects. The quality parameter also specifies iterations, so when testing start with 'draft' quality and go larger as testing progresses.
Can I run Pixray without command line?
Yes, Pixray can be run from within Python code using 'import pixray' and 'pixray.run()'. It also works well in Google Colab notebooks, which is friendly for users unfamiliar with command line. Installation snippets are applicable for Colab users who cannot access the command line.
What hardware does Pixray need?
Pixray works well with 16GB RAM, though working with less RAM is fine (see Working with weaker machines for solutions). For cloud generation, it typically runs on Nvidia T4 GPU hardware, with predictions completed in about seven minutes depending on inputs.
Is Pixray free to use?
Yes, Pixray is open-source and free to self-host using Python, command line, Docker, or Google Colab. Through Replicate, there is initial free usage but after a certain period a credit card is required with charges at $0.0002 per second for predictions, varying by hardware used.
Who is Pixray best for?
Pixray is best for developers, artists, designers, content creators, game artists, and retro aesthetic enthusiasts who want technical control and customization beyond consumer-grade generators. It particularly serves those who need genuine pixel art capabilities and want to fine-tune generation parameters.