* This blog post is a summary of this video.

Installing and Using the Crea AI Diffusion Plugin for Stable Diffusion

Author: AIchemy with XerophayzeTime: 2024-03-23 17:05:01

Table of Contents

Downloading and Installing the Crea AI Diffusion Plugin

The Crea AI Diffusion plugin brings the power of AI image generation with Stable Diffusion models directly into the Crea digital art software. This allows creating AI-generated images in real time by simply scribbling on the canvas. The plugin features the latest AI capabilities thanks to support for model zoos like Automatic1111's CompVis, as well as specialized models like Realistic Visions and Dirac SDK. By following a few simple steps to download and activate the plugin, you can start leveraging AI to enhance your creative workflows in Crea.

The plugin contains everything needed to get AI image generation up and running. The streamlined installation process automatically configures components like Docker containers behind the scenes. Within minutes, the Real Time AI Drawing mode can be activated to paint with AI assistance.

Prerequisites

The only requirement before installing the plugin is having Crea downloaded from the official website and set up on your system. Crea runs on Windows, Mac, and Linux operating systems. Ideally you should install the latest available version to ensure compatibility, but the plugin specifies that Crea 2022.09.24 is the minimum version required.

Installation Steps

The installation process consists of just a few quick steps:

  1. Download the latest plugin archive from the Crea AI Diffusion GitHub repository. This will be a ZIP file containing the plugin code.
  2. Extract the contents of the ZIP file directly into the pyCrea folder located in your Crea user data directory. The pyCrea folder is where Crea loads Python plugins from.
  3. Open Crea and navigate to the Settings. Under the Python Plugin Manager, check the box next to "AI Image Diffusion" to activate the plugin.
  4. Restart Crea to complete the plugin installation process. After starting Crea again, the plugin functionality becomes available. Next we need to configure integration with Stable Diffusion.

Enabling the Plugin in Crea

With the plugin files in place, two quick steps integrate Stable Diffusion:

  1. In Crea's Settings, enable the Docker titled "AI Image Generation". This sets up the backend service powering AI image generation.
  2. In the Docker configuration screen, choose connection settings based on your preferred setup. The option for "Local Server Managed by Crea Plugin" handles everything automatically without needing to run any external software.

Configuring the Plugin Settings

Before starting to generate images, some configuration of the Crea AI Diffusion plugin can help tailor it to your system hardware and creative goals. Options are available to accelerate processing using GPUs, select different Stable Diffusion models, and fine-tune settings that influence the image generation quality and creativity.

Choosing Hardware Acceleration Options

The Crea plugin can leverage NVIDIA GPUs to significantly speed up AI image generation compared to relying solely on CPU processing power. Under the "Hardware" settings, choose "CUDA Nvidia GPU" if running on a system with NVIDIA graphics cards installed. For even better performance, select a specific GPU like "CUDA_0" to use a dedicated graphics card rather than defaulting to the primary display adapter. Consider updating to the latest NVIDIA Game Ready drivers for optimal compatibility with Stable Diffusion. For systems without NVIDIA GPUs, select "Run on CPU" to still use the plugin's capabilities, albeit at reduced performance. The generation of images will be noticeably slower in this mode depending on available CPU power.

Selecting Stable Diffusion Models

The plugin gives control over which Stable Diffusion models get installed to use for image generation. Under the "Models" section, check boxes allow enabling either Stable Diffusion version 1.5 or the XL variant depending on your priorities. Stable Diffusion 1.5 delivers higher image quality and coherence, with less distorted faces and details. It excels at resolution of 512x512 and below. Meanwhile, the XL model handles higher resolutions better starting from 768x768 up to 512x512. XL brings out more abstract, artistic creativity in generated images rather than focusing on accuracy and realism. Additionally, model zoos like Automatic1111 CompVis offer specialized Stable Diffusion models that can be downloaded and configured here as well, such as Juggernaut for anime-style images or Dirac SDK for photo-realism enhancements. Combining Crea's real-time processing with these state-of-the-art diffusion models enables tremendous creative flexibility powered by AI.

Other Configuration Settings

Many other settings allow customizing the image generation processing such as:

  • Sampler - Configure the sampling algorithm use during diffusion, with DPM++ steering output more toward the prompt text
  • Steps - Set the number of denoising steps applied to rendered images
  • CFG Scale - Scale influence of added ControlNet processing for enhanced coherence between scribbles and final output Don't be afraid to experiment with these values to see which settings best bring your AI-assisted artwork visions to life!

Using the Live AI Drawing Feature

With the Crea AI Diffusion plugin enabled and Stable Diffusion models installed, navigating to the "Live" mode activates the mind-blowing capabilities of AI-powered drawing assistance. Essentially, this allows you to scribble on the canvas and watch an AI interpret your rough doodles into beautifully rendered artwork in real time.

Starting a New Canvas

Click the "New" button in Crea to start with a blank document. Access the AI Live Drawing mode through the toolbar button showing a magic paintbrush icon. This enters the interactive generation workflow powered by the configured Stable Diffusion model rather than applying effects after the fact. Next, add a new layer by clicking the "+" button in the layers palette, choosing "Add New Layer", and selecting the desired layer type like "Line Art" or "Soft Edge". Give the layer a name reflecting your envisioned content. Finally, click the magic wand button with sparkles to add this as an AI-responsive layer linked to the diffusion model.

Adding Layers for AI Drawing

The added layer serves as the surface where your scribbles get translated into AI-generated imagery. Think of it like telling an artist friend your ideas and them bringing your vision to life with detailed skill! With the blank layer ready, start sketching out basic shapes and figures with your mouse or drawing tablet. Press play to activate the AI drawing assistant. Incredibly, you'll witness the model interpret and render your raw doodles into cohesive scenes and characters reflecting the layer name as the prompt.

Drawing and Seeing Results in Real Time

It feels nothing short of magical to scribble a few haphazard lines only to watch a photorealistic landscape or elaborate machine materialize from your mouse strokes! The AI continuously analyzes the canvas, following your markings while embellishing details everywhere else. Fine tuning the prompt text steers the model's artistic interpretation toward different themes and moods. Don't be afraid to erase and redraw parts that don't match your vision. The model picks up on the changes and redirects the image accordingly. Toggle between model variations like Stable Diffusion 1.5 and XL to compare different rendering aesthetics. Export your favorite AI-enhanced artwork creations to share with the world!

Tips for Best Results

With an infinite variety of images possible from even the most basic scribbles, mastering AI-assisted art in Crea becomes an captivating journey of imagination. Apply these tips to get the most impressive visuals out of each ephemeral stroke guiding the machine learning model.

Adjusting Control Net Strength

The ControlNet setting determines how strongly the model correlates its outputs back to your specific scribbles instead of freely interpreting strokes however it wants. Lower values give you more surprises as the AI riffs off your drawings into unexpected directions. Higher control tightens coherence toward your markings as dominant guides. Start around 50% control and tweak from there for your desired effect. Too little and your original sketch fades under layers of AI embellishment. Too much overrides the creative explorations you look to AI to provide! Think of yourself as conducting an orchestra of artistic AI agents - provide measured guidance toward visions you find intriguing.

Switching Between Models

The myriad Stable Diffusion model options available all have unique qualities that inspire different images. Play with shifting from Photorealistic model specializations for vivid scenes to Abstract stylizations for avant-garde interpretations of your pencil wanderings! Notice how modulating ControlNet while changing models makes the AI enact various understandings of the same scribbled shapes. Develop a feel for the personalities of each model variation to become a true AI Whisperer guiding procedural generation in desired directions.

Changing Canvas Size

Larger canvas areas allow the AI rendering more room interpret your drawings. Initial small surfaces tend to result in crowded intricate images as complexity compensates for space. Experiment with tweaking the document dimensions bigger to spur more grand, sweeping genre scenes from basic strokes laid down across fresh blank space inviting prismatic manifestation!

Conclusion and Next Steps

Installing the Crea AI Diffusion empowers effortlessly tapping into the creative potential of AI to take personal artwork to new heights. What once required meticulously painting entire scenes can now spring to life automatically from casual sketches in real time thanks to the latest machine learning advancements.

Dive deeper into steering the models by fine-tuning sampling methods, trying new control settings, and blending the strengths of different model specializations. The journey of learning to guide AI image generation in intuitive ways that resonate with your unique inner vision has only just begun! Imagine now what dreams may come as the tools progress even further.

If you found this guide helpful, be sure to check out other videos covering new AI art techniques, updates, and discoveries as they emerge in this rapidly evolving creative frontier. With infinite synergistic possibilities awaiting discovery, the most wondrous AI artistic innovation may spring forth from your own hand next thanks to creative tools that help transcend our inherent human limits of actualization. What will you conjure into being?

FAQ

Q: What are the system requirements for using Crea AI Diffusion?
A: You need a GPU that supports CUDA such as an Nvidia GPU, at least 8GB of VRAM, and 16GB of system RAM for best performance.

Q: Do I need to already have comfy UI installed?
A: No, the plugin can automatically install and manage comfy UI for you when using local server mode.

Q: Why do I get a black canvas when trying to draw?
A: Make sure to turn the control net strength setting up to 100% for live drawing modes like scribble or soft edge.

Q: What's the difference between the 1.5 and XL models?
A: 1.5 models work better for some artistic styles while XL models produce more photorealistic results.

Q: Can I use a prompt when live drawing?
A: Yes, add your prompt text before starting to draw and the AI will incorporate it into the generated imagery.

Q: How do I clear the canvas and start over?
A: You can create a new blank document from the menu to restart your drawing.

Q: Does this plugin allow upscaling images?
A: Yes, there is an upscale tool to increase resolution of parts or all of your final artwork.

Q: Can I animate video with this plugin?
A: Not directly, but you can export each frame as an image and compile into video later.

Q: Is there a way to save brushes or other assets?
A: You can save your Crea documents to continue working on them later or export layers as images.

Q: What is the license and availability of this plugin?
A: Crea AI Diffusion is a free, open-source plugin maintained on GitHub by the developer community.