So AI can "draw" now too. Free & open-source

So AI can "draw" now too. Free & open-source.

I thought of a really easy way to prove that your art is drawn by a human and not made with AI; if you record yourself actually drawing or painting the image — basically if you record a speed paint, then you can prove that you drew it, right? I think this is the easiest way to prove that a piece of art is made by a human and not AI. Well, actually, never mind scratch that because this new AI tool was just released and it is just insane.

Introduction

This new AI can take an input image and recreate the speed paint process of that image. Here’s how it works in detail. You can see that this image was probably generated using stable diffusion, and here the AI recreates the entire speed paint process: from simple lines, coloring, adding details, to showing the entire background and lighting effects. The process appears incredibly realistic and detailed.

Examples

1. Miku Example: This image, generated using stable diffusion, is recreated in the speed paint process. It starts with simple lines, refines them, erases unnecessary parts, fills base colors, adds detailed colors, and finally, adjusts brightness and contrast.
2. Demon Slayer Example: The process naIls the workflow again, going from pencil outlines to shading, refining lines, adding base colors, detailed coloring, and lighting effects.
3. Iterative Drawing Process: The AI tests different outlines for characters, making it look like a human speed painting, not just copying the outlines in one go.
4. More Scenarios: It can handle different types of drawings, from anime to oil paintings to more realistic photos, and even creates the backgrounds similarly to what human artists would do.

Realism and Human-Like Painting Process

While watching these speed paints, you might find it hard to tell if the painting was done by AI or a human. It features realistic shading adjustments, iterative outlines, coloring stages, and lighting effects. For each input image, multiple output processes can be generated, each subtly different but resulting in the same final image.

Tool Features

• Free and Open Source: Runs locally on a computer with a user-friendly Gradio interface.
• Coarse Sketch Extraction: Learn to draw by converting any image into a rough sketch and refning it.
• Coloring Books and Drawing Tutorials: Generate images based on different stages of drawing to create your own tutorials.

Drawing Simulations and Use Cases

• Speed Paint Generation: For images, including anime and more realistic art styles, the AI can generate a convincing drawing process in stages.
• Interpolate Feature: Uses different key frames to create smooth transitions between stages of the painting process.
• Learning Tool: Perfect for learning to draw by studying the generated outlines and coloring stages.

Limitations

• Realistic Photos: The AI struggles with realistic images and might introduce artifacts.
• Complex Scenes: With multiple characters, it might hallucinate and show some imperfections, especially if the image contains intricate UI/UX designs.

Conclusion

In conclusion, PaintsUndo is a powerful AI tool for generating realistic speed paints. However, it does have limitations in handling complex scenes and realistic photos. For artists and enthusiasts, it's an incredible aid for learning and creating realistic drawing tutorials. If you encounter issues during installation or have other related queries, feel free to ask for assistance.

Keywords

• AI Art Generation
• Speed Paint
• Stable Diffusion
• Drawing Process
• Gradio Interface
• Coarse Sketch
• Interpolation
• Learning Tool
• Limitations

FAQ

Q: What is PaintsUndo? A: PaintsUndo is a free and open-source AI tool that can recreate the speed paint process of an image, generating realistic painting stages.

Q: How does this AI tool work? A: It takes an input image and uses deep learning models to recreate the painting process in stages, from simple lines to detailed coloring and lighting.

Q: What are the hardware requirements to run PaintsUndo? A: Ideally, you need a GPU with 16GB of VRAM or more. It does not work well with 8GB of VRAM.

Q: Can it generate speed paints for all types of images? A: It works best for anime and paintings, and struggles with realistic photos or complex scenes with multiple characters.

Q: Can I use PaintsUndo to create teaching materials? A: Yes, it can be used to generate drawing tutorials and coloring books by extracting different stages of the drawing process.

Q: Are there any significant limitations of this tool? A: Yes, it struggles with realistic photos, complex scenes, and might introduce artifacts when dealing with intricate designs.

Q: How can I install PaintsUndo? A: Installation involves downloading and setting up the repository, creating a Conda environment, and running the tool using Python and a Gradio interface.

Q: Can I run PaintsUndo on Colab? A: Currently, the author suggests it might eventually be available as a Colab notebook, but as of now, it needs to be run locally.