Software Engineering Courses (SECourses)•10mo ago

Would it be possible to include some kind of preview while it is being processed? even if it is just

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 SECourses•3/15/25, 4:53 PM

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

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

lets say you generate exactly same video

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

of same seed

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

that turned out bad

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

after restart

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

you get same?_

FFurkan Gözükara SECourses you get same?_

JaviEMOP•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

JaviEMOP•3/15/25, 4:55 PM

I will try it as soon as I can

JaviEMOP•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...

JaviEMOP•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

JaviEMOP•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?

JaviEMOP•3/15/25, 8:49 PM

This settings

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

Furkan Gözükara SECourses•3/15/25, 8:54 PM

you can use google gemini flash

Furkan Gözükara SECourses•3/15/25, 8:54 PM

to generate different angles of person

Furkan Gözükara SECourses•3/15/25, 8:54 PM

train on that

Furkan Gözükara SECourses•3/15/25, 8:55 PM

generate more

Furkan Gözükara SECourses•3/15/25, 8:55 PM

and train againm

GGasby Ok guys. I have a 5090 and tried the WAN 2.1 auto installer app. This is the err...

Furkan Gözükara SECourses•3/15/25, 8:56 PM

you need to install python 3.10 to your system

Furkan Gözükara SECourses•3/15/25, 8:56 PM

then it will work

JJaviEM This settings

Furkan Gözükara SECourses•3/15/25, 8:57 PM

can you try same seed same prompt and image without teacache?

JaviEMOP•3/15/25, 8:57 PM

yes

FFurkan Gözükara SECourses you need to install python 3.10 to your system

Gasby•3/15/25, 8:58 PM

Ok, so python 3.13 isn’t compatible?

GGasby Ok, so python 3.13 isn’t compatible?

Furkan Gözükara SECourses•3/15/25, 8:59 PM

nope

Furkan Gözükara SECourses•3/15/25, 8:59 PM

you dont need to uninstall

Furkan Gözükara SECourses•3/15/25, 8:59 PM

just install 3.10 my installer will handle

FFurkan Gözükara SECourses can you try same seed same prompt and image without teacache?

JaviEMOP•3/15/25, 8:59 PM

processing. I will tell you in 1/1.5 hours

FFurkan Gözükara SECourses just install 3.10 my installer will handle

Gasby•3/15/25, 9:01 PM

Thank you

JJaviEM processing. I will tell you in 1/1.5 hours

Tomfoolery•3/15/25, 9:01 PM

You dont need 50 steps to test, nor 81 frames.

Drop it to 20 steps and 33 frames for testing purposes. will save tons of time.

TTomfoolery You dont need 50 steps to test, nor 81 frames. Drop it to 20 steps and 33 fram...

JaviEMOP•3/15/25, 9:03 PM

I tried 20 steps once and I got a completely bugged video

JaviEMOP•3/15/25, 9:03 PM

I'll test 20 steps again next time

Furkan Gözükara SECourses•3/15/25, 9:04 PM

i can test faster if it is not NSFW. send me output and image

FFurkan Gözükara SECourses i can test faster if it is not NSFW. send me output and image

JaviEMOP•3/15/25, 9:04 PM

Don't worry, it's already in progress

JaviEMOP•3/15/25, 9:05 PM

I am in no hurry

JJaviEM I tried 20 steps once and I got a completely bugged video

Tomfoolery•3/15/25, 9:12 PM

Hm, you definitely have something weird going on. This is 1.3b model, 25 steps, 49 frames for "A cinematic air-view tracking shot of a train waving through a forest, peaceful and serene."

Not a masterpiece, but a quick way to generate outputs to test prompts / functionality. This one took 104 seconds.