Co-Creating with AI

Exploring and Integrating Stable Diffusion Models in the Creative Process of Poster Design.


03    Experiments



03 .01     Text Prompts

In the text-to-image generation process, testing prompts between the base Stable Diffusion v1.5 model and personalised models was important to understand the impact of fine-tuning. Initial generic prompts were used for these models to generate images.

Stable Diffusion text prompts.
Stable Diffusion, v1.5, txt2img



Model 1 text prompts. Stable Diffusion, Model 1, txt2img



In the fine-tuning process of the AI models, a trigger word was added. It is a unique identifier used in prompts to call the images used during fine-tuning.

Here are some images with the trigger word.  Visual differences can be seen  when compared to the previous images.

Model 1 text prompts with trigger word.
Stable Diffusion, Model 1, txt2img, triggered



Model 2 text prompts. Stable Diffusion, Model 2, txt2img



Model 2 text prompts with trigger word. Stable Diffusion, Model 2, txt2img, triggered



Model 3 text prompts. Stable Diffusion, Model 3, txt2img



Model 3 text prompts with trigger word. Stable Diffusion, Model 3, txt2img, triggered



Model 4 text prompts. Stable Diffusion, Model 4, txt2img



Model 4 text prompts with trigger word. Stable Diffusion, Model 4, txt2img, triggered



Model 5 text prompts. Stable Diffusion, Model 5, txt2img


© 2024 BASEL