Model Types

Learn about different types of Open Sourced Stable Diffusion Models

About [6 bit], [4.5 bit], and [4-bit] Model Versions

These versions are three times lighter than the originals, saving significant memory. They can run on devices with limited RAM, including iPhone 12-13, iPad with A15 chip, and older Macs that may struggle with the original models.

Please note that there may be minor differences in result quality compared to the original versions.


Standard Models

These are variant models based on Stable Diffusion 1.5, 2, and 2.1. They can produce amazing results while maintaining a lighter weight and faster generation speeds. Most of them support ControlNet for additional input and control.

Don’t be fooled by the fact that they are based on SD 1.5; these models have been heavily trained by the open-source AI community and are capable of producing images of very high quality and complexity.


SDXL Models

Stable Diffusion XL is the latest image generation model developed by Stability AI. It is designed to produce more photorealistic outputs with greater detail compared to previous SD models. It can create descriptive images with shorter prompts, generate text within images, and offers improved image composition and face generation, resulting in stunning visuals and highly realistic aesthetics. ‍ SDXL generates images at 1024x1024 pixels by default, but enabling the upscaler will result in 2048x2048 pixel images, matching the behavior of other models when the upscaler is activated. ‍ It's important to note that the XL models are also the heaviest in terms of computational demands, requiring substantial processing power and memory. As a result, it may lead to slower generation times.


SDXL Turbo Models

SDXL Turbo achieves state-of-the-art performance with a new distillation technology, enabling single-step image generation with unprecedented quality, reducing the required step count from 50 to just 1-3 steps.


LCM Models

Latent Consistency Models (LCM) && SDXL-TURBO are a way to decrease the number of steps required to generate an image with Stable Diffusion (or SDXL) by distilling the original model into another version that requires fewer steps (4 to 8 instead of the original 25 to 50). This allows you to get more images and multiple variants in a fraction of the time. Make sure to select the LCM scheduler when using these models. ‍ LCM models typically yield better results with 2 to 8 steps and a low Guidance Scale of 1-2. When you select the LCM scheduler, the Steps and Guidance Controls' value scale is automatically adjusted to allow for a more fine-grained value setup. Because of the tight range of values, small changes in Step count and Guidance Scale can have dramatic effects on the quality of the generated image. However, generating images with LCM models is a quick process and therefore convenient for previewing the effects of your settings. ‍ Tip: LCM models generate excellent results with the help of a Guide Image and Control Net. If you generate an image that you think could be better, set it as a Guide Image with a strength value between 15-50, shuffle the seed, and regenerate it to quickly enhance it.


Need help? Ask in our Discord!

For any questions regarding the use of Sogni, tips, suggestions, bugs, sharing art, discussing AI and life, or simply hanging out... Join our Discord community! Join Sogni Discord ✨

Last updated