[Stable Diffusion] Prompt Sharing and Learning Thread

[modine2021]

anyone know how to fix this? of of a sudden Comfy decided it dont want to render no more. but it worked before. i get this error using Wan sampler AND Ksampler now::



WanVideoSampler (happens with KSampler also)
Command '['M:\\AI\\SwarmUI\\dlbackend\\python_embeded\\Lib\\site-packages\\triton\\runtime\\tcc\\tcc.exe', 'X:\\temp\\tmpmaemgsyk\\__triton_launcher.c', '-O3', '-shared', '-Wno-psabi', '-o', 'X:\\temp\\tmpmaemgsyk\\__triton_launcher.cp313-win_amd64.pyd', '-fPIC', '-D_Py_USE_GCC_BUILTIN_ATOMICS', '-lcuda', '-lpython313', '-LM:\\AI\\SwarmUI\\dlbackend\\python_embeded\\Lib\\site-packages\\triton\\backends\\nvidia\\lib', '-LM:\\AI\\SwarmUI\\dlbackend\\python_embeded\\Lib\\site-packages\\triton\\backends\\nvidia\\lib\\x64', '-IM:\\AI\\SwarmUI\\dlbackend\\python_embeded\\Lib\\site-packages\\triton\\backends\\nvidia\\include', '-IM:\\AI\\SwarmUI\\dlbackend\\python_embeded\\Lib\\site-packages\\triton\\backends\\nvidia\\include', '-IX:\\temp\\tmpmaemgsyk', '-IM:\\AI\\SwarmUI\\dlbackend\\python_embeded\\Include']' returned non-zero exit status 1.

 

[modine2021]

anyone know how to fix this? of of a sudden Comfy decided it dont want to render no more. but it worked before. i get this error using Wan sampler AND Ksampler now::



WanVideoSampler (happens with KSampler also)
Command '['M:\\AI\\SwarmUI\\dlbackend\\python_embeded\\Lib\\site-packages\\triton\\runtime\\tcc\\tcc.exe', 'X:\\temp\\tmpmaemgsyk\\__triton_launcher.c', '-O3', '-shared', '-Wno-psabi', '-o', 'X:\\temp\\tmpmaemgsyk\\__triton_launcher.cp313-win_amd64.pyd', '-fPIC', '-D_Py_USE_GCC_BUILTIN_ATOMICS', '-lcuda', '-lpython313', '-LM:\\AI\\SwarmUI\\dlbackend\\python_embeded\\Lib\\site-packages\\triton\\backends\\nvidia\\lib', '-LM:\\AI\\SwarmUI\\dlbackend\\python_embeded\\Lib\\site-packages\\triton\\backends\\nvidia\\lib\\x64', '-IM:\\AI\\SwarmUI\\dlbackend\\python_embeded\\Lib\\site-packages\\triton\\backends\\nvidia\\include', '-IM:\\AI\\SwarmUI\\dlbackend\\python_embeded\\Lib\\site-packages\\triton\\backends\\nvidia\\include', '-IX:\\temp\\tmpmaemgsyk', '-IM:\\AI\\SwarmUI\\dlbackend\\python_embeded\\Include']' returned non-zero exit status 1.

EDIT: never mind. fixed by reinstalling triton

 

[JhonLui]

EDIT: never mind. fixed by reinstalling triton

I've noticed that anytime there is an update involving Torch and/or Torch components, you have to reinstall triton, as it will be deactivated.

 

[rogue_69]

Appreciate this thread, thanks to all who have contributed. I'm releasing a fairly massive Twine game soon-ish that uses AI images, as a necessary evil, despite my dislike of most all things AI. This thread definitely helped some.

You're setting yourself up for failure if you "dislike most things AI" while also needing to use AI as a necessary evil. Either use it and embrace it, or don't use it. Otherwise you'll feel shame about it, and won't really enjoy your results no matter how things turn out. Not picking on you, just trying to give some life advice.

 

[rogue_69]

Wan 2.2 starting off with the image and then using "hard cut" to get to the first actual frame of the video. Then used the second to last frame of each video to start a new one. I feel like the character consistency was acceptable.

 

[HinsdaleDays]

You're setting yourself up for failure if you "dislike most things AI" while also needing to use AI as a necessary evil. Either use it and embrace it, or don't use it. Otherwise you'll feel shame about it, and won't really enjoy your results no matter how things turn out. Not picking on you, just trying to give some life advice.

I get what you're saying, but we're talking about AI porn here. I think talking about feeling "shame" around this is a little bit in the rear view mirror

 

[JhonLui]

I get what you're saying, but we're talking about AI porn here. I think talking about feeling "shame" around this is a little bit in the rear view mirror

Just get a good model and try to do something nice.
The quality around is so low that sometimes I think a blue screen would be more exciting.

 

[rogue_69]

I get what you're saying, but we're talking about AI porn here. I think talking about feeling "shame" around this is a little bit in the rear view mirror

I think you're misunderstanding. I don't think anyone should feel shame about it. My point is that if you are going to use it, don't be ashamed. Embrace it. Or if you're against using it, then don't use it.

 

[Sepheyer]

Are there video workflows which cut down the rendering time like 10x?

I've been out of the scene for a quick second, but it felt like I caught a glimpse of a youtube video talking about a breakthrough with regards to rendering speed.

While I love Wan 2.2, waiting for 20 minutes per video eventually destroyed my enthusiasm.

 

[JhonLui]

Are there video workflows which cut down the rendering time like 10x?

I've been out of the scene for a quick second, but it felt like I caught a glimpse of a youtube video talking about a breakthrough with regards to rendering speed.

While I love Wan 2.2, waiting for 20 minutes per video eventually destroyed my enthusiasm.

I had a chat on that matter with a guy on Stability Matrix discord.
It seems that Wan 2.2 in Comfy has reached Hunyuan in Framepack F1 in therms of speed, using the latest GGUF model and an accelerator Lora 4Steps + Sage Attention.
The downside (I've red around) is the memory handling upon generations above 4 seconds.
I don't remember what your specs are, but if you have 16GB Vram and above 32 Ram, Wan 2.2 seems to be the better option.
In my case (8GB Vram and 32GB Ram) Framepack F1 is still the best option: 20-25 mins for a 12 secs video, even if Hunyuan is a lower quality model.

 

[devilkkw]

This

You must be registered to see the links

is made with wan2.2 (image via zit, audio with acestep).
Size is 1024x768 in wan, at 24fps, 90frame. done with 6gb vram without any patch or lora. Last version of ComfyUI seem have better memory management.
The car part, is done by using

You must be registered to see the links

, audio is done with acestep, with modified sampler.

 

[Sepheyer]

This

You must be registered to see the links

is made with wan2.2 (image via zit, audio with acestep).
Size is 1024x768 in wan, at 24fps, 90frame. done with 6gb vram without any patch or lora. Last version of ComfyUI seem have better memory management.
The car part, is done by using

You must be registered to see the links

, audio is done with acestep, with modified sampler.

Gives me "page not found"

 

[JhonLui]

Gives me "page not found"

Exactly

By the way.. do me a favour.
I'm a shit at prompting and I want to make a definitive trial on the model.
Please give me a prompt and a subject (I'll use ItoV) and a lenght of video, so that I can see what comes out and in what time, as it seems the latest Nvidia driver has finally done some good (!?!).
For yesterday I wanted to try a new option and turned out I made a 6sec video in under 9 mins, but still have hard times in getting a girl undressing....

If not it's ok, I don't mean to be rude.

 

[Sepheyer]

Exactly

By the way.. do me a favour.
I'm a shit at prompting and I want to make a definitive trial on the model.
Please give me a prompt and a subject (I'll use ItoV) and a lenght of video, so that I can see what comes out and in what time, as it seems the latest Nvidia driver has finally done some good (!?!).
For yesterday I wanted to try a new option and turned out I made a 6sec video in under 9 mins, but still have hard times in getting a girl undressing....

If not it's ok, I don't mean to be rude.

I don't really have any experience with the video prompting - in fact the few videos I made were meh, downright bad I'd even say

Here is a prompt from ComfyUI's templates Text-2-Image:

"A skier glides over fresh snow, kicking up large clouds of snow as he turns. Accelerating gradually down the slope, the camera moves smoothly alongside."

And here's what it looks like:

 

[JhonLui]

I don't really have any experience with the video prompting - in fact the few videos I made were meh, downright bad I'd even say

Here is a prompt from ComfyUI's templates Text-2-Image:

"A skier glides over fresh snow, kicking up large clouds of snow as he turns. Accelerating gradually down the slope, the camera moves smoothly alongside."

And here's what it looks like:

View attachment 5558473

Meanwhile I was working on this, LOL


But thanks, I'll give it a try on a 12 secs video with standard settings 640x480 than upscaled.

 

[devilkkw]

Gives me "page not found"

I checked and it work, i don't know why you get error. Can you find it in my civ profile?
Another question: is "show your AI art" thread gone?

 

[JhonLui]

Ok, sorry for the delay, but I had guests for dinner...

Nope.. still fast, but slow, yet the result is not bad (don't post the upscaled because is too big..)
The second one would be me Wiiiiiiiii!

 

[Sepheyer]

I checked and it work, i don't know why you get error. Can you find it in my civ profile?
Another question: is "show your AI art" thread gone?

 

[Sepheyer]

is "show your AI art" thread gone?

Yes. Been for a while now. There was something - an incident or a warning or some such - that made F95 review their policy either on younger humans or on dogs loving humans making the policy more restrictive. And unless those things are in a game they can't be posted as an image.

Maybe it was that Steam saga that shrapnelled into F95. But one day it was: rule change - bonk! And then: threads gone - bonk!

 

[devilkkw]

Don't understand why the civitai won't show, actually is down and can't check,
This is the

You must be registered to see the links

on reddit.
Thank you for explain. Is there a location where you share your creation?