Welcome Stable Diffusion 3 :)
Update 06/14
I created a new workflow today specifically for portraits of a person. The workflow highlights the strengths of SD3 and tries to compensate for its weaknesses. I found that SD3 eyes look very good, but the skin textures do not. Therefore, in this workflow, the faces are detected and the eyes are subtracted, so only the skin is improved while keeping the beautiful SD3 eyes. However, this only works for close-ups. Additionally, I added an automated color grading; the Kodachrome LUT file is in the .zip folder of the workflow.
For images that are not close-up portraits, I recommend the 5-person face detailer workflow.
12/06 Update - 3:30 (CET)
The weights have been released, and I have improved an error in the workflow, now it works. Have fun with SD3 and leave some reviews with your SD3 images.
19.04.24 Update.
I used the official Stability Ai Comfyui nodes and added the Creative Upscaler, which also runs through the API.
Since today, you can use Stable Diffusion 3 via the Stability API. Unfortunately, the facial quality, as expected from stability, is not so convincing. Therefore, I've built a workflow that enhances it. Up to 5 faces in an image can be improved.
U can get your API Key here:
Stability AI - Developer Platform
Description
FAQ
Comments (8)
Can you use SD1.5 checkpoints to fix the faces?
"Cannot execute because a node is missing the class_type property.: Node ID '#51'"
What goes there?
For some reason, my results are looking super burnt when using this workflow or any that use the CLIPTextEncodeSD3 node, but when using the regular CLIPTextEncode, it's fine. Did you run into this problem as well?
What about Lykon’s SD3 model?
Thank you for the workflow! Great job! I really appreciate it! The detail_seg is kinda mess up the nose eh? Any idea which setting should I tweak it?
Yeah! Keep em coming!
where did you get the SD3 vae?
@MakeThemComeAliveAIArt
https://civitai.com/models/511494/sd3-vae
Looks like we don't have an active mirror for this file right now.
CivArchive is a community-maintained index — we catalog mirrors that volunteers upload to HuggingFace, torrents, and other public hosts. Looks like no one has uploaded a copy of this file yet.
Some files do get recovered over time through contributions. If you're looking for this one, feel free to ask in Discord, or help preserve it if you have a copy.