Software Engineering Courses (SECourses)•10mo ago

0.15 is pretty much max i believe. above will make faster

Furkan Gözükara SECoursesOP•3/15/25, 1:57 PM

but quality may degrade

Furkan Gözükara SECoursesOP•3/15/25, 1:57 PM

try 0.2

Furkan Gözükara SECoursesOP•3/15/25, 1:57 PM

over 100% speed probably

FFurkan Gözükara SECourses 0.15 is pretty much max i believe. above will make faster

AiInfluence•3/15/25, 2:03 PM

and for gpu L40?

AAiInfluence and for gpu L40?

Furkan Gözükara SECoursesOP•3/15/25, 2:09 PM

same you can try to load all params in gpu

Furkan Gözükara SECoursesOP•3/15/25, 2:09 PM

if works

Furkan Gözükara SECoursesOP•3/15/25, 2:09 PM

someone said there is a banding problem with RIFE

Furkan Gözükara SECoursesOP•3/15/25, 2:09 PM

is he here?

JaviEM•3/15/25, 3:12 PM

Can the RAM bug cause quality degradation? I've made about 3 or 4 videos in a row with awful results, each one worse than the previous one. I restarted the application and it generated a much more natural video.

JJaviEM Can the RAM bug cause quality degradation? I've made about 3 or 4 videos in a ro...

Furkan Gözükara SECoursesOP•3/15/25, 3:40 PM

i doubt that. actually i will now try several consquent ones lets see what happens

Furkan Gözükara SECoursesOP•3/15/25, 3:43 PM

testing multi line

Furkan Gözükara SECoursesOP•3/15/25, 3:43 PM

JaviEM•3/15/25, 4:00 PM

in my experience long promps have always given me bad results

Furkan Gözükara SECoursesOP•3/15/25, 4:43 PM

@JaviEM all looking normal

Furkan Gözükara SECoursesOP•3/15/25, 4:43 PM

you doing image to video?

Furkan Gözükara SECoursesOP•3/15/25, 4:44 PM

00152.txt978B

00154.txt982B

00153.txt978B

FFurkan Gözükara SECourses you doing image to video?

JaviEM•3/15/25, 4:50 PM

Yes. I am getting very mixed results. Some videos are perfect, some are awful.

JaviEM•3/15/25, 4:51 PM

It's not funny to wait 1 hour for a horrible video with an extreme hallucination

JaviEM•3/15/25, 4:52 PM

Would it be possible to include some kind of preview while it is being processed? even if it is just a random image of the video to see if everything is correct or if it is better to cancel it.

JJaviEM Would it be possible to include some kind of preview while it is being processed...

Furkan Gözükara SECoursesOP•3/15/25, 4:53 PM

i am not sure since vae decoding happens at the end.

Furkan Gözükara SECoursesOP•3/15/25, 4:54 PM

lets say you generate exactly same video

Furkan Gözükara SECoursesOP•3/15/25, 4:54 PM

of same seed

Furkan Gözükara SECoursesOP•3/15/25, 4:54 PM

that turned out bad

Furkan Gözükara SECoursesOP•3/15/25, 4:54 PM

after restart

Furkan Gözükara SECoursesOP•3/15/25, 4:54 PM

you get same?_

FFurkan Gözükara SECourses you get same?_

JaviEM•3/15/25, 4:55 PM

I would need to test it, but I would lose 1 hour without being able to use the PC knowing that the result could be bad

JaviEM•3/15/25, 4:55 PM

I will try it as soon as I can

JaviEM•3/15/25, 4:56 PM

I miss when I thought that the 5 minutes it took Supir to rescale an image was a long time

daroth•3/15/25, 5:52 PM

its so much harder to experiment with img2vid when one try takes 8 minutes

masood•3/15/25, 6:04 PM

@Dr. Furkan Gözükara i am trying to create a virtual influencer, ( i want to combine multiple people and make one new person ) but the problem i am facing is consistency, the output person doesnt stay consistent it chnges with every pic generated. I am using flux dev with swarmui, any suggestion?

JJaviEM Would it be possible to include some kind of preview while it is being processed...

Tomfoolery•3/15/25, 6:07 PM

Swarm and Comfy both have preview options for generating videos, so I am sure its possible to integrate. Now - how complicated or resource-intensive the integration... thats a different question.

Generally I agree though, I would also love to see a preview feature to ensure that the random number gods didnt bless the current generation with nonsense.

TTomfoolery Swarm and Comfy both have preview options for generating videos, so I am sure it...

JaviEM•3/15/25, 6:08 PM

I lost too many hours on bugged videos, so preview would be an huge improvement in this app

Tomfoolery•3/15/25, 6:10 PM

Im also tinkering with adding the ability to integrate <random:X, Y, Z, etc> tags into the prompt handler. So this way you can set it to run overnight with something like:

A beautiful <random:sunset, sunrise, rainy day> over the city.

And it might generate:

A beautiful sunset over the city.
A beautiful sunrise over the city.
A beautiful rainy day over the city.

Tomfoolery•3/15/25, 6:11 PM

Its something else that Swarm does that I love, lets me write a fairly complicated prompt with various options and variations, set a whole bunch of generations, and walk away for a few hours or overnight

Mmasood @Dr. Furkan Gözükara i am trying to create a virtual influencer, ( i want to com...

Pelolisu•3/15/25, 7:09 PM

did you try this ? https://huggingface.co/spaces/primecai/diffusion-self-distillation

Diffusion Self Distillation - a Hugging Face Space by primecai

JaviEM•3/15/25, 7:52 PM

another hour lost. First generation after start wan 2.1, so it's not affected by ram bug.

Mmasood @Dr. Furkan Gözükara i am trying to create a virtual influencer, ( i want to com...

GomezBro•3/15/25, 8:26 PM

I created two loras of two different women. I then use the weights of both loras in comfyui and set them like .50 and .75 and generate a new woman using both loras, this creates a new woman that looks like both but different. This is how i create a new woman and its consistent, i also mass generate images and just select the good ones

GomezBro•3/15/25, 8:27 PM

U can do this for your inquiry

JJaviEM another hour lost. First generation after start wan 2.1, so it's not affected by...

Tomfoolery•3/15/25, 8:29 PM

Which model are you using, and what are your teacache settings?

Tomfoolery•3/15/25, 8:30 PM

and steps/ CFG/ lora?

Gasby•3/15/25, 8:44 PM

Ok guys. I have a 5090 and tried the WAN 2.1 auto installer app. This is the error I’m getting:

flash_attn-2.7.4.post1-cp310-win_amd64.whl is not a supported wheel on this platform

deepspeed-0.16.4-cp310-win_amd64.whl is not a supported wheel on this platform

TTomfoolery Which model are you using, and what are your teacache settings?

JaviEM•3/15/25, 8:49 PM

This settings