Today, we’re diving into a service developed using Scade.pro by one of our clients. This service simplifies the creation of graphic illustrations from text, using a chain of AI.
Here’s a step-by-step breakdown of how it turns a simple user input into a vivid, detailed image.
The interface of the service requires the user to fill out just three fields to get the image:
- What to illustrate.
- What to consider.
- Ratio.
But under the hood, more than one neural network is working to produce the image.
- The user’s request is received in the workspace on the Scade.pro platform.
- A translation node converts the user’s request into English and passes it to the idea generator.
- The idea generator powered by an internal framework houses an internal framework that enhances the creativity of the request. enhances the creativity of the request. It breaks down the task, delves into each component, identifies the best solution, and refines it into a final concept. While we can’t disclose all the secrets of the prompt due to a trade secret, our framework is based on the principles outlined in ‘Patterns for Creative Thinking’ by Christian Kohls.
- The refined concept is then turned into a prompt that instructs the AI on how to describe the illustration and select from over 30 graphic styles. This step ensures that the neural network can accurately capture the mood of the image.
- To further define the illustration, short cues describing the style, such as 2D flat, are added to the prompt.
- Additionally, a negative prompt is explicitly written to prevent the generation of distorted images, ensuring quality
-
- The comprehensive prompt is processed by the Fooocus node model, which generates the final image.
- This image is then saved on the server, allowing users to access it anytime through their history.
- A unique feature of this process is the enhancement of facial features. If the image includes a face, it’s passed through our retrained Stable Diffusion model. This model enhances the facial features, which clarifies and naturalizes facial features, adding a touch of realism to the generated illustration.
This entire process eliminates the need for users to craft detailed requests manually, making it easier than ever to create custom graphics. It’s a testament to how neural networks can simplify complex tasks, providing tools that are as innovative as they are accessible.
Register at Scade.pro and gain access to over 1,500 AI models with intuitive drag-and-drop flow builder. After you register, we’ll drop 200 credits to your account.