Stable diffusion thread

[tokong]

Can use this Easy Diffusion (which actually is Stable Diffusion).

Download the Windows installer. (Yes, no need the hassles of "pip install" on Ubuntu or WLS (Windows Subsystem for Linux).) Free. CPU and GPU flexible. On CPU it will take a bit of time, but I find it quite ok, because I am just playing around with it during my free time.

If you have a gaming computer, you will really enjoy it a lot.

https://easydiffusion.github.io/

 

[doogyhatts]

Stability.ai announces Stable Video 3D.
It is a major improvement over their previous Stable Zero123.
https://stability.ai/news/introducing-stable-video-3d

 

[doogyhatts]

Stability.ai releases SD3-Turbo paper.
Only 4 steps.
Not released yet.

Spoiler

 

[doogyhatts]

SD3 will be the last T2I model from SAI as the company is falling apart.

Spoiler

 

[zheng]

SD3 will be the last T2I model from SAI as the company is falling apart.

Spoiler

gg no more sd

 

[doogyhatts]

ByteDance releases AnimateDiff-Lightning!
https://huggingface.co/ByteDance/AnimateDiff-Lightning/tree/main

Now I can generate the facial expressions animations much faster using AnimateDiff at 53 frames (discard 22 frames).

 

[doogyhatts]

 

[doogyhatts]

 

[zheng]

SD3 is close?

 

[doogyhatts]

SD3 is still being worked on.

Spoiler

 

[doogyhatts]

SD3 has entered early beta!

 

[zheng]

SD3 has entered early beta!

anywhere can test?

 

[doogyhatts]

anywhere can test?

I haven't receive the invitation via email.

 

[doogyhatts]

So I found out that the quality output from AnimateDiff Lightning is actually lower for realistic animation, and it has a faint watermark.
It should be more suitable for cartoon visuals or non-realistic ones.

 

[order66]

very long never play with AI liao

are people still using SD1.5 or all use XL / newer versions already?

 

[doogyhatts]

very long never play with AI liao

are people still using SD1.5 or all use XL / newer versions already?

Most people are still using SD1.5, some use XL. All of us are waiting for SD3.

Just FYI.
The Ai_vlog_girl YT channel got demonetised because he kept doing under-boob.
So semi-lewd AI content cannot be monetised anymore on YT.
Although some much smaller channels doing bikini photobooks are already monetised.

Recently, I have found a sub-genre on ASMR where real women do face up close recording.
My most recent face up close video using AI girls got at least 11K impressions in 12 hours after release.

 

[order66]

Most people are still using SD1.5, some use XL. All of us are waiting for SD3.

Just FYI.
The Ai_vlog_girl YT channel got demonetised because he kept doing under-boob.
So semi-lewd AI content cannot be monetised anymore on YT.
Although some much smaller channels doing bikini photobooks are already monetised.

Recently, I have found a sub-genre on ASMR where real women do face up close recording.
My most recent face up close video using AI girls got at least 11K impressions in 12 hours after release.

i just went to see see look look at civitai
almost all the new models are based on XL, dont see many new models for SD1.5
think maybe i should take the chance to install XL

 

[zheng]

i just went to see see look look at civitai
almost all the new models are based on XL, dont see many new models for SD1.5
think maybe i should take the chance to install XL

after I switch to SDXL + turbo already... didnt really go back to 1.5 anymore

 

[doogyhatts]

i just went to see see look look at civitai
almost all the new models are based on XL, dont see many new models for SD1.5
think maybe i should take the chance to install XL

Are you already using ComfyUI? or are you still using A1111?

I used Stability Matrix to help me install Fooocus, Forge and ComfyUI.
I only use SDXL in Fooocus but with SDXL-Lightning 8 steps.

 

[order66]

Are you already using ComfyUI? or are you still using A1111?

I used Stability Matrix to help me install Fooocus, Forge and ComfyUI.
I only use SDXL in Fooocus but with SDXL-Lightning 8 steps.

still using A1111

maybe i'll wait abit until SD3 public release, then just play around with that instead of XL