Krita, generate images with artificial intelligence on your PC

Fell is open source software for digital graphics creation, illustration and digital painting. It is primarily designed for digital artists, illustrators, cartoonists and animators. Equipped with a clear and intuitive user interface, Krita is designed to be easily usable even by beginners.

The application integrates a wide range of brushes and painting tools, also supports vector graphics, allows you to work with layers, integrates animation tools and supports a wide range of graphics tablets. Once the work is completed, users can export the image to various common formats such as JPEG, PNG or PSD for sharing or printing.

Krita AI Diffusion: the main features of the plugin to create works of art

The authors of the project Krita AI Broadcastwhose source is publicly available on GitHub, have developed a plugin which allows you to generate images efficiently and intuitively using artificial intelligence.

The software Krita AI Broadcast, developed in Python, bases its operation on the well-known Stable Diffusion generative model. The model uses diffusion algorithms stochastic to generate new images or to modify and extend existing images in a coherent and convincing way.

Generating visual content with artificial intelligence

By installing the plugin Krita AI Broadcastit’s possible marking an area present in the image opened in the application using the Krita selection tools (Inpaint). This way you can remove or replace the contents existing.

The function Outpaintinstead, expands the canvas, or Krita’s “canvas” on which you can create your own work. User can select a blank area by letting Krita AI Broadcast automatically fills it with capable content blend perfectly with the one already present in the image.

The instrument Generate allows you to create new images starting from a natural language description or using pre-existing images. The plugin can use and leverage two specific versions or variations of the template Stable Diffusion, each with different characteristics and capabilities. Users can choose between SD1.5 e SDXL to adapt image generation to their specific needs or obtain different results based on the model used.

Edit images with generative AI

This video shows the potential from the plugin Krita AI Broadcast.

Content optimization created with Krita AI Diffusion

The plugin Krita AI Broadcast also integrates one slider which allows you to refine content rather than completely replace it. It is also not necessary to ask too many questions about the resolution of the image you wish to obtain: the application, in fact, is able to select the most appropriate number of pixels vertically and horizontally by applying any scaling.

The intelligent upscaling technique allows you to increase the size of any image without quality loss. In particular, it is easy to transform low-resolution images into graphical content 4K or 8K format without exhausting the RAM available for data processing.

Image upscaling with artificial intelligence

Workflow management

Since generating images can take some time tempo depending on the hardware configuration in use, the plugin Krita AI Broadcast it allows to line up and cancel jobs while other processing is carried out on the image.

The chronology Furthermore, integrated into the application, it gives you the opportunity to preview the results and conveniently browse the contents as they are generated.

The ability to store and reuse preset predefined i.e. “recycled” configurations between several different graphic projects, finally allows us to avoid replicating the same interventions on similar images processed at different times.

How to install Krita

Krita is available in versions for Windows, macOS and Linux. To use the plugin satisfactorily Krita AI BroadcastHowever, it is highly recommended to have a system equipped with a powerful graphics card (at least 6 GB of VRAM on board). It is certainly possible to carry out the calculations relying only on the CPU but the time required to complete them will often be unsustainable.

As a further option, you can rent one GPU cloud paying a variable amount for the time strictly necessary to generate the images. In this case, the processing is totally independent of the system configuration in use locally and moves to the cloud. To proceed with using Krita AI Diffusion on the cloud, just follow the instructions on the page Rent a Cloud GPU then download and launch the Docker container also containing the plugin.

Krita, generate AI images on the cloud

If you choose to install Krita locally, you need to download the Krita AI Diffusion plugin Zip file. From the menu Tools, Scripts of Krita it is therefore necessary to click on the item Import Python extension from file. You must then select the Zip file of Krita AI Broadcast without unzipping it then reply Yes when the message “Enable extensions now?”

Krita AI plugin import

Install and configure the Krita AI Diffusion plugin

The first step to create an image with AI consists of clicking on the icon New image thus specifying the size of the canvas in pixels. To continue you must click Crea so up Settings, Docking areas and finally up AI Image Generation. Finally, you can finally click the button Configure in the box AI Image Generation which appears at the bottom right of the Krita window.

Image generation with AI

Now you need to click on the button Install so you can let the AI ​​plugin automatically download and set up all the necessary software components: Python, PyTorch, Stable Diffusion, and so on. Approximately 10 GB of disk space is required.

Generating an image with artificial intelligence

To generate an image with AI and the Stable Diffusion generative model, we suggest starting with something first small size: start, for example, with a new 512×512 pixel image. By clicking on the button Configure placed in the box AI Image Generator (already seen previously), you must select the button Launch.

When the green message “Server running – Connected“, you can click OK at the bottom right.

In the text entry box, you can at this point describe the image in words that you want to achieve. It is important to provide, at this stage, as many details as possible in clear, simple and precise language.

The system generates more images simultaneously: by clicking on the boxes at the bottom right, it is possible to switch from one graphic content to another. You can also navigate between preset called Cinematic Photo e Digital Artwork to explore two possible examples of achievable effects. With a click on the gear symbol, it is possible customize each stile.

Create an image from a simple sketch

With the plugin Krita AI Broadcast you can even generate a professional image from a basic sketch. Even the simple drawing of a preschooler can lead to the generation of amazing results.

To proceed in this direction, you need to click on the button again Configure then activate the box ControlNet Scribble below Control extensions and finally click the button Install. After clicking OK, you can sketch a drawing with the subject you want to obtain.

How to generate images with AI starting from a drawing

By clicking on the small button in the figure, choosing Scribble from the menu on the left, describing the content of the drawing and finally clicking on Generate, you will obtain a quality image, like the one represented in the example.

Image generated with Stable Diffusion from drawing

By changing simple details (words and adjectives) in the text description, it is possible to create the perfect image by proceeding with a series of subsequent refinements. It’s interesting to check what happens when you select Digital Artwork and renewing the generation of the image.



Please enter your comment!
Please enter your name here