Software Engineering Courses (SECourses)•3y ago

For two weeks I have tried all the variables that I found in tutorials for dreambooth without good r

For two weeks I have tried all the variables that I found in tutorials for dreambooth without good results.
And I don't see many realistic Lora in civitai either, the only one I found that has some good ones is this https://civitai.com/user/dogu_cat
I don't know what I'm missing but it seems to be quite complicated to achieve reliable results
Your examples are all great but I can't replicate that similarity level

dogu_cat Creator Profile | Civitai

Average Rating: 4.9 (149), Models Uploaded: 31, Followers: 453, Total Likes Received: 4.9k, Total Downloads Received: 41k.

FFractal For two weeks I have tried all the variables that I found in tutorials for dream...

Furkan Gözükara SECourses•3/19/23, 4:46 PM

what you want to achieve?

FFurkan Gözükara SECourses what you want to achieve?

FractalOP•3/19/23, 4:59 PM

replicate a face in some realistic way
I always end up with results that look like the person but not exactly

FractalOP•3/19/23, 5:00 PM

I caption the image set, and I have tried using regularization images too

FractalOP•3/19/23, 5:00 PM

doesn't change much

fqgedecot•3/19/23, 5:28 PM

Hey Everyone I Really Very Urgently Need Help With My App.
Is There Anyone Just Anyone Who Could Help Me With Integrating Stable Diffusion Or Any API That Can Simply Take 3-5 Images Of Users And Create It's AI Avatar, Sort Of Like Lensa.
Please Anyone Just Any One Help Me With It It's Extremely Really Very Important.

fqgedecot•3/19/23, 5:28 PM

Even If Anyone One Of You Know A Person Who Could Help Me Please Connect Me To Them Just Please Help Me Guys.

fqgedecot•3/19/23, 5:30 PM

@Furkan Gözükara SECourses Can You Help Me With IT???.

FFractal replicate a face in some realistic way I always end up with results that look li...

Furkan Gözükara SECourses•3/19/23, 5:49 PM

you want photo realism right?

Furkan Gözükara SECourses•3/19/23, 5:49 PM

512px is not just sufficient enough

Ffqgedecot @Furkan Gözükara SECourses Can You Help Me With IT???.

Furkan Gözükara SECourses•3/19/23, 5:50 PM

people are paying thousands of dollars to developers for this task

fqgedecot•3/19/23, 5:51 PM

I Can Offer A Percentage Of Equity In My Brand, Or Could Pay You After I Recieve Sales.

fqgedecot•3/19/23, 5:52 PM

If You Could Help Me Make It Or Know Somebody Who Could Make It???.

Ffqgedecot If You Could Help Me Make It Or Know Somebody Who Could Make It???.

Furkan Gözükara SECourses•3/19/23, 5:52 PM

i doubt that you can find anyone with this way

fqgedecot•3/19/23, 5:53 PM

There Must Be Some Way.

fqgedecot•3/19/23, 5:53 PM

Lensa Made It Possible.

fqgedecot•3/19/23, 5:53 PM

Remini Followed Suite, There Is Something They Know That I Don't. What Possibly It Could Be.

Furkan Gözükara SECourses•3/19/23, 6:01 PM

of course it is possible

Furkan Gözükara SECourses•3/19/23, 6:02 PM

it just cost money and time to get it done

Major Epidemic•3/19/23, 7:38 PM

Anyone else got weird workflows?
I find the best likeness I can get to my subject is if I train using automatic1111 Dreambooth then use Kohya to extract a LoRA from it. It’s nearly always way more accurate than training a Lora by itself in Kohya. I can also generate a bunch of checkpoints and extract all of them to find the one with the least overfitting.

Furkan Gözükara SECourses•3/19/23, 7:47 PM

i only used web ui dreambooth extensin and textual inversion so far

Zono50•3/19/23, 9:12 PM

Trained a model for like 6 hours today in auto1111 dreambooth, the sample images looked just like my subject, but when i go to txt2image, and put in the lora name, it looks nothing like my subject

ZZono50 Trained a model for like 6 hours today in auto1111 dreambooth, the sample images...

Furkan Gözükara SECourses•3/19/23, 9:12 PM

something must be off not making sense

Furkan Gözükara SECourses•3/19/23, 9:12 PM

i am still processing new video

Zono50•3/19/23, 9:13 PM

yeah it's weird, the instance name like photo of ohwx isn't used unless i convert it to a ckpt right? i should just be able to put in the lora tag <> and it should work

ZZono50 yeah it's weird, the instance name like photo of ohwx isn't used unless i conver...

Furkan Gözükara SECourses•3/19/23, 9:14 PM

i think now it supports lora activation

Furkan Gözükara SECourses•3/19/23, 9:14 PM

are you acivating your loras?

Zono50•3/19/23, 9:14 PM

how do i activate it?

Furkan Gözükara SECourses•3/19/23, 9:14 PM

i didnt test newest version yet

Furkan Gözükara SECourses•3/19/23, 9:14 PM

you need to type lora file name

Zono50•3/19/23, 9:14 PM

i mean i'm going into the additional networks, going to lora tab, and clicking on the pt file

Furkan Gözükara SECourses•3/19/23, 9:14 PM

if you generate ckpt

Furkan Gözükara SECourses•3/19/23, 9:14 PM

then it is ofc activated

Furkan Gözükara SECourses•3/19/23, 9:14 PM

with ohwx man word

Zono50•3/19/23, 9:15 PM

so it's putting the <lora model name> for txt2img, but like i said, it looks nothing like my subject, even when i use the 1.5-pruned that it was trained on

Zono50•3/19/23, 9:15 PM

i checked lora and lora extended

Zono50•3/19/23, 9:15 PM

did about 300 epochs

Furkan Gözükara SECourses•3/19/23, 9:15 PM

i will make a new lora video hopefully soon

Zono50•3/19/23, 9:16 PM

it's just strange cause every 25 epochs it was printing out images that looked like my subject, than it finishes, then looks nothing like my model

Zono50•3/19/23, 9:18 PM

i'll print out the settings i used here in a sec as soon as my venv folder regenerates

Furkan Gözükara SECourses•3/19/23, 9:37 PM

Zono50•3/19/23, 9:37 PM

Got this weird error while trying to convert my lora to a checkpoint Exception compiling checkpoint: The size of tensor a (4) must match the size of tensor b (320) at non-singleton dimension 1

Major Epidemic•3/19/23, 9:37 PM

I also have a weird one where if I use the Lora I trained in comfyui it looks perfect. If I use it in automatic1111 it displays some gibberish with exactly the same prompt.

Zono50•3/19/23, 9:38 PM

interesting

Zono50•3/19/23, 9:38 PM

maybe i'll generate an image of my lora model on comfy and see what i get

Major Epidemic•3/19/23, 9:39 PM

Yeah try it. I still haven’t worked out why it is doing it yet but my head was aching so will look tomorrow.

Zono50•3/19/23, 9:40 PM

how do i load a lora model on comfyui

AlxHsInf•3/19/23, 9:42 PM

@Furkan Gözükara SECourses It may be helpful to you if you create channels for the different GPU Cards people have, like 20series, 30series, 40 series, 16series, that way people can confirm if your updates worked on their graphics cards, and also people with those graphic cards can go see if they should update or wait...

AlxHsInf•3/19/23, 9:42 PM

https://www.nvidia.com/en-us/geforce/graphics-cards/

NVIDIA

NVIDIA GeForce Graphics Cards

GeForce RTX 40, 30, 20 series & GTX 16 series.

ZZono50 how do i load a lora model on comfyui

Major Epidemic•3/19/23, 9:52 PM

Bring it in after the model but before clip. I can’t remember off the top of my head but I think the node is under loader and then Lora.

Zono50•3/19/23, 9:53 PM

yeah i don't get it, i make a model and when i activate it, it's completely different

For two weeks I have tried all the variables that I found in tutorials for dreambooth without good r

Similar Threads