so i dnt even need to worry abt 4 gb files per model
so i dnt even need to worry abt 4 gb files per model
Rhere are the results from my subject LoRA. These are ai
Fyou need extra extension
RUsing some images from a studio shoot idid
Fto generate lower size files
Fwith dreambooth extension
Fi plan to make a video but havent done yet
Fi had been waiting them to update and i think they did
Fso i can make a new video
Fmaybe tomorrow
Rbut I havent managed to get good body outfit mix yet
Fi think easier way would be
Fgenerate person with outfit
Fand inpaint face
Rthe inpaint would be using the subject LoRA only. Right. Ok
Fi got to sleep see you tomorrow hopefully
Fso many videos i need to make so little time :d
Rthank you so much! Get your well deserved sleep Dr
RSo are you simply using your Face LoRA InPainted?
Pstill a lot of testing to do. at least I didn't turn into a robot or something halfway through the video. style and structure remained consistent throughout. so it's progress.
Pwicked awesome results
KLook awesome 
CIf all my Dreambooth checkpoint outputs are grainy, what does this indicate?
CAll outputs on txt2img with the trained model are grainy as well. Running 23GB Vram.
C
CDoes this have to do with the VAE?
Pcould be your model is overbaked
Ptoo much training
Pthis image def looks overbaked
KThat's a super cool idea, Dr. Furkan! Can we also use it for checking styles and objects? We're kinda stuck because SD can churn out way more images than we can check manually. It doesn't have to be perfect, but if it could rank each image from 1 to 10, it would save us a ton of time and effort
CAt all sample steps, training images are blurry:
C500, 600, 700
C10 High quality training images with diverse lighting and angles, all cropped to 512
Pyou should run a x/y plot with your checkpoints
CWill do one sec
CX/Y is not working for some reason but here is 1000 steps:
C
C800 steps
C
C600 steps
C
C@PhillieD Does the quality look off to you as well?
P600looks best
CDo your training sample images look as grainy as mine?
Pdon't quite understand what you're asking
CI feel like I can be achieving much higher quality, feels like there is no-vae distortion.
CThese are some of my outputs from the dreambooth/samples folder
CAll the sample images look like this, at every ckpt
Pso you haven't tried generating new images using the checkpoints you created? these are just the images generated during training?
CAll those "forest guardian" images are new generations at each checkpoint
