Late to the party as usual, I finally got around fine-tuning my own images with Stable Diffusion XL 1.0 on a free-tier Google Colab using LoRA. Despite many tutorials being available so far, it took a bit of trial and error to get optimal results. I used the SDXL weights from StabilityAI on images of my cat Thor and fine-tuned them using the Dreambooth method which requires only a few (how many is a few?) images of the subject. I chose SDXL because I wanted my photos to have a high resolution (1024x1024). There seems to be conflicting opinions on... [Read More]

To be honest, I was never interested in trying Midjourney - sure you can create cool pics, but what could I use them for? After all, I live in a small apartment, and our walls are already full of photos of cats and paintings that we received as gifts or thrifted. [Read More]

Stable diffusion is an exciting field in GenAI for generating all sorts of images, videos, but also to enhance your own picture. There are many many flavours of stable diffusion models (fine-tuned for a specific tasks, or one specific training images) and versions (v1, v2, XL, etc.). For this tutorial I will be using Segmind Stable Diffusion (SSD), a 50% smaller and faster stable diffusion that generates pics as good as the OG stable diffusion from Stability AI. Aditionally, I will use the HuggingFace implementation of SSD in a Python Notebook on Google Colab and prompt the SSD model plus... [Read More]