Header

OdysseyXL!

So for the past few weeks I thought to myself what projects I can do to add to my portfolio. Naturally I went to ChatGPT and asked it this prompt:

If you were a hiring manager at a leading AI company, what would be some portfolio project would you like to see?

GPT-4o gave me a lot of really good options but there was one particular one that stood out to me. It was to create my own diffusers model. After about an hour of research I decided to stick with SDXL Base 1.0

I went with SDXL for many reasons. Some include:

Open Weights
Already a really good model
Stability.ai's community licence

and many more which I can't remember of the top of my head

I collected data by using hand-collected data by going outside and taking pictures and added them to other images from free image providers (For example Unsplash)

After fine-tuning OdysseyXL, I generated some images to compare the fine-tune with it's based model:

001

As we can see. The fine-tune can now better generate faces due to the dataset having high-quality images of faces. OdysseyXL can also add more brighter and natural tones to its subjects and it has better prompt adherence compared to SDXL

You can use OdysseyXL aswell with this code snippet!

from diffusers import DiffusionPipeline
 
pipe = DiffusionPipeline.from_pretrained("Spestly/OdysseyXL-v1")
 
prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k"
image = pipe(prompt).images[0]

So in the end, I learnt alot about SDXL and I now have my own fine-tune of it!