Software Engineering Courses (SECourses)•10mo ago

@Dr. Furkan Gözükara i am trying to create a virtual influencer, ( i want to combine multiple people

@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

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?

JaviEM•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?

JaviEM•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...

JaviEM•3/15/25, 9:03 PM

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

JaviEM•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

JaviEM•3/15/25, 9:04 PM

Don't worry, it's already in progress

JaviEM•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.

TTomfoolery Hm, you definitely have something weird going on. This is 1.3b model, 25 steps, ...

Furkan Gözükara SECourses•3/15/25, 9:12 PM

really good

Furkan Gözükara SECourses•3/15/25, 9:12 PM

i am gonna add <random:X, Y, Z, etc> now

TTomfoolery Hm, you definitely have something weird going on. This is 1.3b model, 25 steps, ...

JaviEM•3/15/25, 9:14 PM

Does it work the same with 14b? I tested with 14b model. I have no problems with 1.3b

JJaviEM Does it work the same with 14b? I tested with 14b model. I have no problems with...

Tomfoolery•3/15/25, 9:22 PM

sure does, takes a bit longer to generate (10 minutes or so at 20 steps, 33 frames), but still works.

Tomfoolery•3/15/25, 9:23 PM

ill generate the same prompt now

Tomfoolery•3/15/25, 9:26 PM

you can also 'scale down' the 720p model and it handles it wonderfully, which speeds up generation times for testing

Tomfoolery•3/15/25, 9:28 PM

10 minutes at full resolution for 20 steps, 33 frames, teacache at 0.15

Tomfoolery•3/15/25, 9:29 PM

Again - not the best output in the world, but definitely lets you test things without having to spend an hour per generation.

Tomfoolery•3/15/25, 9:29 PM

if you wanna save the headache of waiting each time.

Tomfoolery•3/15/25, 9:34 PM

This is same inputs (different seed) as before, but using 16:9 low size. 3.5 minutes to generate. This oue actually turned out really nice

TTomfoolery This is same inputs (different seed) as before, but using 16:9 low size. 3.5 min...

Furkan Gözükara SECourses•3/15/25, 9:37 PM

very cool

JaviEM•3/15/25, 9:52 PM

@Tomfoolerywhat gpu and ram do you have?

Tomfoolery•3/15/25, 9:55 PM

3090 Ti, 64 gb RAM.

Tomfoolery•3/15/25, 9:56 PM

But the basic premise that lower steps/frame/size to speed up testing runs should work on all hardware.

Furkan Gözükara SECourses•3/15/25, 10:05 PM

i am about to add random word sytnax

Furkan Gözükara SECourses•3/15/25, 10:05 PM

and also fixing auto scale

@Dr. Furkan Gözükara i am trying to create a virtual influencer, ( i want to combine multiple people

Similar Threads

Similar Threads

Similar Threads