A Step Closer to Stable Diffusion XL with SDXL 0.9

A dark gothic manor during heavy rain with people walking towards it, a person carrying umbrella, artwork created with SDXL 0.9
DarkGothicManor_HeavyRain_ZeroNine, a synthography art created with SDXL 0.9 (courtesy: Author)

In synthography, specifically text-to-image model. Stable Diffusion is arguably one of the most well-known (alongside DALL-E).

Since its release in 2022, thousands of synthography artworks have been created with Stable Diffusion.

It seems that many people are not too happy with Stable Diffusion 2.x due to perceived lower performance and the necessity to use negative prompts for a good artwork.

Improvements are welcome, especially for emerging software like Stable Diffusion and here why people are eager for the final release of SDXL (Stable Diffusion XL). SDXL is a next generation iteration with better capabilities compared to the “old” Stable Diffusion (1.x and 2.x) series.

Previous synthography posts by Vulcan in this blog are featuring artwork created with SDXL Beta, an earlier development version of SDXL.

Here are some touted improvements of SDXL 0.9 compared to earlier SDXL Beta, according to StabilityAI:

The key driver of this advancement in composition for SDXL 0.9 is its significant increase in parameter count (the sum of all the weights and biases in the neural network that the model is trained on) over the beta version. 

SDXL 0.9 has one of the largest parameter counts of any open source image model, boasting a 3.5B parameter base model and a 6.6B parameter model ensemble pipeline (the final output is created by running on two models and aggregating the results). The second stage model of the pipeline is used to add finer details to the generated output of the first stage.

StabilityAI SDXL 0.9 Release Announcement

A step closer… and so, on 22 June 2023, SDXL 0.9 was released to the public (aimed for research purposes).

A gothic girl in a cyberpunk world artwork created with SDXL 0.9
Gothic Meets Cyberpunk, a synthography artwork created with SDXL 0.9 at ClipDrop (courtesy: Author)

You can try this by using DreamStudio, ClipDrop (platforms by StabilityAI), and third-party platform like NightCafe Studio (Vulcan prefers this).

Have fun creating!

  • M. Ananda Laksmana (1712501202)
  • Nugraha Dwiky (1712502598)
  • Panji Adipramono (1912501317)

Leave a Reply

Your email address will not be published. Required fields are marked *