Stable Diffusion Version

Stable Diffusion Versions

There are 3 notable versions of Stable Diffusion:

• Version 1.4/1.5 - The original open source model released by Stability AI, these two weights are what were released to the public in 2022. A quite popular version, there are thousands of weights, textural inversions, and LoRA's available based on this design.
• Version 2.x - The next version, released by Stability AI, is an improvement on Stable Diffusion. Not as widely adopted, but the model is different and incompatible with weights from 1.4/1.5
• SDXL - The latest version, released by Stability AI, is a far different model than the prior two and has improved even more.

Differences between the legacy model and contemporary model

What changed between the two versions of Stable Diffusion were:

The Text Encoder model

• SD1.4/1.5 uses OpenAI's CLIP Encoder model
• SD2.x uses the OPENClip Encoder model

The Unet architecture

• The Unet is located inside the diffusion model, and what changed was the architecture of the Unet.

Legacy vs Contemporary Mode

In MetalDiffusion, if you're using the TensorFlow Render Engine, then you will need to explicitly state which version of Stable Diffusion you are using.

Note: You can not use SDXL on the TensorFlow Render Engine

Stable Diffusion 2.0 (SD) uses two different models than SD 1.4 and 1.5. It's important to select the correct mode when creating with your model(s).

Under Settings, select your model. Then select Use Legacy Stable Diffusion if you are using any SD models made prior to the release of SD 2.0.