Software Engineering Courses (SECourses)•6mo ago

Hey, I have been finding resources and documents for text to video for wan with images being dataset

Hey, I have been finding resources and documents for text to video for wan with images being dataset or image to video with videos being dataset, but I was wondering if there is a way to train lora with images and use first and last image as reference and generate a video based on the reference images? do you happen to know any models which lets you train on images and then use that lora with reference images?

Uunderdog Hey, I have been finding resources and documents for text to video for wan with ...

Furkan Gözükara SECourses•7/27/25, 5:10 PM

i dont know such model

Furkan Gözükara SECourses•7/27/25, 5:10 PM

but wan 2.1 can be trained from images to generate videos

007•7/27/25, 5:52 PM

Hey guys, so I'm really impressed with the fine tune accuracy with regards to my model's face. it looks great. but it still has a bit of AI look to it. and I know the model is trained on photorealism.

my goal is it make it look more candid, like it was shot on an iPhone or a digital cam. I've used loras with varying strengths and it completely changes the faces on the model. I've tried different resolutions but it doesnt help much.

007•7/27/25, 5:52 PM

should I try and combine the fine tune with a realism fine tune or would that totally mess it up?

0007 Hey guys, so I'm really impressed with the fine tune accuracy with regards to my...

Furkan Gözükara SECourses•7/27/25, 10:58 PM

more realism is obtainerd with

Furkan Gözükara SECourses•7/27/25, 10:59 PM

more images + more training

Furkan Gözükara SECourses•7/27/25, 10:59 PM

but if you use a more realism oriented fine tuned base flux model

Furkan Gözükara SECourses•7/27/25, 10:59 PM

it may also help

FFurkan Gözükara SECourses but if you use a more realism oriented fine tuned base flux model

007•7/27/25, 11:17 PM

I actually just tried that but it keeps giving me an error. Specifically I'm using ultrarealfinetunev4 as the base model and its saying

KeyError: 'time_embed.0.weight'

Chatgpt is saying:
"You're running into a common issue: the UltraRealFineTune .safetensors file is not a full Stable Diffusion checkpoint — it's missing key components, particularly UNet time embedding weights, which causes this error:

KeyError: 'time_embed.0.weight'"

007•7/27/25, 11:18 PM

so assuming I cannot use that model do you have any reccomendations?

0007 I actually just tried that but it keeps giving me an error. Specifically I'm usi...

Furkan Gözükara SECourses•7/28/25, 12:42 AM

chat gpt or alikes will know nothing

0007 I actually just tried that but it keeps giving me an error. Specifically I'm usi...

Furkan Gözükara SECourses•7/28/25, 12:42 AM

are you training stable diffusion

Furkan Gözükara SECourses•7/28/25, 12:42 AM

or flux?

FFurkan Gözükara SECourses are you training stable diffusion

007•7/28/25, 12:46 AM

in this case I did everything the same as your Flux dreambooth finetune video except I changed the base model, so flux im assuming? I noticed that it will work when the "flux1" box is checked but not otherwise

0007 in this case I did everything the same as your Flux dreambooth finetune video ex...

Furkan Gözükara SECourses•7/28/25, 12:59 AM

yes it is flux

Furkan Gözükara SECourses•7/28/25, 12:59 AM

UltraRealFineTune is flux model? how many gb?

FFurkan Gözükara SECourses UltraRealFineTune is flux model? how many gb?

007•7/28/25, 1:02 AM

https://civitai.com/models/978314?modelVersionId=1413133

UltraReal Fine-Tune - v4 | Flux Checkpoint | Civitai

V4 Alright, so what’s new in this version? I cranked up the aesthetic dial, added more diversity in ages, and improved how it handles Asian feature...

007•7/28/25, 1:02 AM

it's over 22gb so I figured it would be the full model

007•7/28/25, 1:05 AM

should I check the flux1 box and train or will that mess it up?

0007 https://civitai.com/models/978314?modelVersionId=1413133

Furkan Gözükara SECourses•7/28/25, 1:09 AM

yes this should work

Furkan Gözükara SECourses•7/28/25, 1:09 AM

Full Model fp16 (22.17 GB)

Furkan Gözükara SECourses•7/28/25, 1:09 AM

if not working you should open an issue thread for kohya to fix

Furkan Gözükara SECourses•7/28/25, 1:09 AM

to support model loading format

FFurkan Gözükara SECourses yes this should work

007•7/28/25, 1:52 AM

I'll go ahead and train it with the "flux1" checked I'm assuming that means im telling kohya its a flux1 model. we will see what happens.

007•7/28/25, 1:52 AM

thanks for your help

0007 I'll go ahead and train it with the "flux1" checked I'm assuming that means im t...

Furkan Gözükara SECourses•7/28/25, 10:14 AM

you are welcome

FFurkan Gözükara SECourses but wan 2.1 can be trained from images to generate videos

underdogOP•7/28/25, 12:23 PM

yes, but can we use reference images/start frame and end frame?
Wan 2.2 is out

Uunderdog yes, but can we use reference images/start frame and end frame? Wan 2.2 is out ...

Furkan Gözükara SECourses•7/28/25, 12:28 PM

ye following it. you can discuss in #flux-stable-diffusion-such

FFurkan Gözükara SECourses ye following it. you can discuss in #flux-stable-diffusion-such

filthy•7/28/25, 6:07 PM

hello sorry to ask but what training parameters should i use for training 56 image model

filthy•7/28/25, 6:07 PM

also if im out of space currently on massed compute how can i add gpu without losing all of my data?

Ffilthy hello sorry to ask but what training parameters should i use for training 56 ima...

Furkan Gözükara SECourses•7/28/25, 9:54 PM

same as our config

Furkan Gözükara SECourses•7/28/25, 9:54 PM

still do up to 200 epoch if you have time

Ffilthy also if im out of space currently on massed compute how can i add gpu without lo...

Furkan Gözükara SECourses•7/28/25, 9:54 PM

sadly you cant as far as i know @NicB

Shura490•7/29/25, 7:23 AM

@Dr. Furkan Gözükara If I want to do a full finetune of Flux is this the best tool? https://github.com/bmaltais/kohya_ss/ Do you have a sample configuration? Thanks!

GitHub

GitHub - bmaltais/kohya_ss

Contribute to bmaltais/kohya_ss development by creating an account on GitHub.

SShura490 @Dr. Furkan Gözükara If I want to do a full finetune of Flux is this the best to...

Furkan Gözükara SECourses•7/29/25, 9:11 AM

kohya

Furkan Gözükara SECourses•7/29/25, 9:11 AM

yes we have

Furkan Gözükara SECourses•7/29/25, 9:11 AM

https://youtu.be/FvpWy1x5etM

YouTubeSECourses

FLUX Full Fine-Tuning / DreamBooth Training Master Tutorial for Win...

If you want to train FLUX with maximum possible quality, this is the tutorial looking for. In this comprehensive tutorial, you will learn how to install Kohya GUI and use it to fully Fine-Tune / DreamBooth FLUX model. After that how to use SwarmUI to compare generated checkpoints / models and find the very best one to generate most amazing image...

Furkan Gözükara SECourses•7/29/25, 9:12 AM

up to date configs here both lora and dreambooth / fine tuning : https://www.patreon.com/posts/click-to-open-post-used-in-tutorial-112099700

Patreon

Kohya FLUX Fine Tuning (Full Checkpoints) Training Full Tutorial Fo...

Get more from SECourses: FLUX, Tutorials, Guides, Resources, Training, Scripts on Patreon

Jamiewhelton•7/31/25, 8:08 AM

Need good data augmentation tips for making a more realistic dataset of an AI person. Specifically more skin detail on the face

Jamiewhelton•7/31/25, 8:17 AM

best EPOC for 39 images?

Jamiewhelton•7/31/25, 8:18 AM

my images are between 1536x1536 and 2152 x 2152 res

is that an issue? flux will downscale, but is there a way I can get more out of these higher rep images in my fine-tuned model?

JJamiewhelton best EPOC for 39 images?

Furkan Gözükara SECourses•7/31/25, 8:53 AM

do at least 200 and compare

JJamiewhelton my images are between 1536x1536 and 2152 x 2152 res is that an issue? flux will...

Furkan Gözükara SECourses•7/31/25, 8:53 AM

1024 working best

Furkan Gözükara SECourses•7/31/25, 8:54 AM

bigger res not helping much

Furkan Gözükara SECourses•7/31/25, 8:54 AM

you can agument your data and add zoomed in face images etc

skyrrr•7/31/25, 4:25 PM

Hey, can we try a full finetune of flux.krea using the same method ? Or not yet ?

Sskyrrr Hey, can we try a full finetune of flux.krea using the same method ? Or not yet ...

Furkan Gözükara SECourses•7/31/25, 5:08 PM

probably

Furkan Gözükara SECourses•7/31/25, 5:08 PM

i need to test

skyrrr•8/1/25, 1:00 PM

Does anyone know if I can expect much faster iteration speed (48GB config file) with a H100 rather than a L40S for a full dreambooth finetune ? Thanks

Sskyrrr Does anyone know if I can expect much faster iteration speed (48GB config file) ...

Furkan Gözükara SECourses•8/1/25, 1:34 PM

a bit faster yes

Hey, I have been finding resources and documents for text to video for wan with images being dataset

Similar Threads