Quickstart Guide to Stable Diffusion 3

Quickstart Guides, SD3

Last Updated	Changes
6/12/2024	First Version
6/17/2024	Licensing Update

? Civitai Stable Diffusion 3 Ban – Updated! 7/22/2024

We’ve rescinded the Civitai SD3 Ban. For more details, please read our announcement, here.

Please be aware, we’ve enacted a temporary Stable Diffusion 3 (SD3) resource ban on Civitai.com while we seek clarification from Stability AI and our legal team on the terms of the SD3 License. New SD3 based resources will not be accepted at this time, and existing resources have been temporarily Archived.

What is Stable Diffusion 3?

Stable Diffusion 3 (SD3) is a text-to-image model, released by Stability AI on June 12th, 2024. This next evolution in their model line follows the groundbreaking SDXL, which was launched around the same time last year.

You can read our Pre-Release Overview here.

For more technical details, please refer to the Research paper.

Where do I get it?

You can downl oa d all the SD3 safetensors, Text Encoders, and example ComfyUI workflows from Civitai, here.

SD3 Medium (4.2 GB)
SD3 Medium Incl Clips (5.8 GB)
SD3 Medium Incl Clips T5XXLFP8 (10.6 GB)

What do I download? IMPORTANT! READ ME!

ComfyUI has full SD3 support! Make sure to update to the most recent version!

~~At time of writing (10:00am EST 6/12/2024), Automatic1111 WebUI did not have support for SD3. We expect support to follow within the next week.~~

Update 7/22/2024 – Limited support available for SD3 on Automatic1111 WebUI dev b ranc h.

There are two SD3 checkpoints which contain text encoders: SD3 Medium Incl Clips and SD3 Medium Incl Clips_t5xxlfp8

These models can be used like any regular checkpoint in ComfyUI.

sd3_medium.safetensors includes the MMDiT and VAE weights but does not include any text encoders. You’ll have to download those separately – see below.

sd3_medium_incl_clips_t5xxlfp8.safetensors contains all necessary weights, including fp8 version of the T5XXL text encoder, offering a balance between quality and resource requirements.

sd3_medium_incl_clips.safetensors includes all necessary weights except for the T5XXL text encoder. It requires minimal resources, but the model’s performance will differ without the T5XXL text encoder.

Make sure to put either sd3_medium_incl_clips.safetensors or sd3_medium_incl_clips_t5xxlfp8.safetensors in your ComfyUI/models/checkpoints directory.

The SD3 Medium base file does not contain text encoder/CLIP weights so you must load them separately to use that file.

Download the text encoder weights from Civitai and put them in your ComfyUI/models/clip/ directory. sd3_medium.safetensors should be put in your ComfyUI/models/checkpoints/ directory.

How do I make images?

You can download pre-made ComfyUI workflows here;

License Considerations

SD3 has license considerations! It Stable Diffusion 3 Medium is released under the Stability AI Non-Commercial Research Community License.

The model is free to use for non-commercial purposes such as academic research.

This model is not available for commercial use without a separate commercial license from Stability. Please visit https://stability.ai/license to learn more.

Training SD3

We haven’t seen much talk from Stability regarding training against SD3, however, we do know that to; “… expect SD3-Medium training requirements to be similar and slightly lower than SDXL.” S o urce.

We also know, from the SD3 announcement email, that the model is “Capable of absorbing nuanced details from small datasets, making it perfect for customization and creativity.“

Any Other Details?

A Stability AI employee announced, via Reddit, that they will be releasing the other SD3 model versions – for free – as they’re finished training, including Small (1B parameter), Large (4B parameter), and Huge (8B Parameter) versions.

EDUCATION