Stability AI released SD XL 0.9 a month ago as beta release and AI art community loved the results that came with it. Now Stability AI has released the fully fledged version of SDXL as SDXL 1.0.
SDXL is now available for download for all the users. Since SDXL is already producing some great images compared to Stable diffusion 1.5 model, it will be interesting to see the how fine tuned models will perform.
The models trained on SD 1.5 were a sudden hit and community kept working on sd 1.5 despite the release of SD 2.x models as those were heavily criticised due to the amount of censorship and the results that it produced.
SDXL 0.9 was met with positive reviews but was that because of the lower expectation after 2.x models or those were genuine praises, only time will tell.
Table of Contents
SDXL 1.0 access
You can access stable diffusion xl 1.0 on various platforms
If you use automatic1111 or any WebUI like comfy the models are available for download on huggingface. both base and refiner are linked. You will need both of these models to get good results.
However if you don’t have a powerful GPU like me, you can still access the Stable Diffusion XL 1.0 on Clipdrop. Clipdrop offers unlimited creations for free, but be vary the waiting time is a lot, but hey its free.
Another website where you can access is Nightcafe. Currently they’re offering 10 free generations and its 1 credit per generation if you use their pro feature.
Comparison of SDXL vs previous generations
SDXL vs SD 1.5
SDXL 1.0 VS SD 2.1
SDXL 1.0 VS SDXL 0.9
CONCLUSION
From the above comparisons i found out that sdxl is more coherent and sticks to the prompt very efficiently than its precursors. Its leaps and bound above Stable diffusion 1.5 which was released 1 year ago and far better than Stable diffusion 2.1.
I am yet to make an informed decision whether it is better than SDXL 0.9 or not as both the images are so close to each other.
What do you think? Let me know in the comments