Advertisement: Click here to learn how to Generate Art From Text
I was inspired by some collage images that I saw on Midjourney, which mimicked newspaper clippings that are glued into shapes or characters. This inspired me to create a SDXL LoRA style (Low Rank Adaptation), which would emulate this in some degree.
As I have documented, the process for creating LoRAs by using Runpod.io on this postI started off by creating an image dataset with captions.
Then, I ran Runpod and created the different LoRA files with different Network Rank settings. The Network Rank parameter is used to train the LoRA. However, the higher the number, the larger the LoRA file.
SDXL 1.0 is my base model, and I have used it to ensure the best possible learning.
I have created some unique pieces of art using the LoRAs that I selected. I would like to share with you one of these LoRAs so that you can experiment, enjoy and create your own artwork using this LoRA.
Here is an example of a comparison between keyword alone and keyword + LoRA, at various CFG Scales (5, 7, and 9). You get a good idea of how much the LoRA for Newspaper Collage influences the generated image.
What is included in this LoRA package?
After purchase, you can download the zip file and extract it. The contents include:
- newspaper-collage.safetensors – LoRA file (435 Mb)
- license.txt – License agreement
- Newspaper Collage LoRA.pdf – Installation Instructions and sample images with their prompts
- You must have Stable Diffusion enabled on your PC/Mac or Cloud GPU (I am using Runpod).
- SDXL 1.0 Base Model must be downloaded
- Download the newspaper-collage.safetensors file and place it in your models/Lora folder
- For Automatic1111: place the file under “SD-Web-UImodelsLora” folder
- For ComfyUI: place the file under “ComfyUImodelsloras” folder
- Use trigger word “newspaper collage style” in your prompt
Below is a collection created using SDXL Base Model 1.0 and Newspaper Collage Lora.