Lykon
@Lykon4072
Followers
3K
Following
1K
Media
581
Statuses
2K
Been off the internet for a while. Spent months living in Japan, focusing on work, self-improvement, and actually enjoying life offline. Happiest I’ve been in years. That said… I’ve got some projects brewing. Stay tuned.
11
1
66
Morning Diffusion world! We've just added support for Stable Diffusion 3.5 Large ControlNet models by Stability AI: Blur, Canny, and Depth—each packed with 8 billion parameters! 🤯 Check out these amazing results you can now run on your local GPU! ✨
6
56
411
We heard you want Control. 🎮 So, today we are adding new capabilities to Stable Diffusion 3.5 Large with the release of three ControlNets: Blur, Canny, and Depth. The models are free for both commercial and non-commercial use under the permissive Stability AI Community
14
68
366
@StabilityAI Thanks to @dango233max @Lykon4072 @virushuo for building SD3.5 and @cubiq @multimodalart for early feedback❤️
0
1
10
To all fellow #SD35 tuners: If you have experienced difficulties in baking details into the model, please try the patched Kohya sd-scripts sd3 branch: * The default timesteps scheduling (Uniform) should be more SFT friendly and attend to details * More adjustable parameters now
4
10
36
New "Animagine" model called Animaestro announced for #SD35L and/or #SD35M by @CagliostroLab
https://t.co/U5sNfGHmJz
1
14
46
We just added an example refiner/upscaler workflow using SD3.5L+M. It's a super basic one that can be improved using ERSGAN too and adding a text truncate node to use a single prompt. Build from here: https://t.co/lRC8OBgEcw
1
6
46
Make sure to update @ComfyUI to get the best out of SD3.5 Large and Medium. SLG by @dango233max was just added (mostly tested with Medium but should work on Large too changing the default settings). We recommend sampling with Skip Layer Guidance Generated with L refined with M
0
2
34
Another meme prompt by @cubiq
trying to recreate this "picture-in-picture" effect with just prompting in SD35L. considering no lora, no controlnet, no inpainting.... it's impressive
1
0
15
The world is not quite what it seems This image was created with SD3.5 Large from @StabilityAI no post processing no compositing didn't even use any negative prompts SD 3.5 is open source and runs in 16g or less vram everything you need to run it or develop with it is here:
0
5
9
This took some trial and error, but the fact itself that it worked from prompt alone is kind of crazy.
25
23
505
Awesome thread
1
0
9