Transform Your Photos into Cinematic Sci-Fi Images with Stable Diffusion and Dreambooth!

[@turgay9956]

is it possible to add/upload custom background images and retrain the model with these backgrounds instead of prompt generated backgrounds?

[@paulonteri]

I love the patience and the level of detail!

[@diptipande5786]

please tell me what u used for face cropping ?

[@fransoavirtuality]

Wouah, Thx for the tut!!!!! Where did you get the class files with all the strange faces...

[@BrentLeBlancCG]

Sure! This is the channel I learned alot of what I did here : www.youtube.com/@Aitrepreneur Here is the video: kzbin.info/www/bejne/q6TKoqCHZbKkbck Person Training Set: github.com/aitrepreneur/REGULARIZATION-IMAGES-SD

[@SantoValentino]

Thanks! I’m training now. Break time. SO…. If I want to use a checkpoint model from civitai, how do I use their model with the one I’m training now? Should I learn how to checkpoint merge? OR can I type (example person1:1) in a downloaded model prompt? Say I download reliberate or whatever is popular or use the base 1.5 and type in a prompt, it won’t use my face unless I add a hyper network or lora. Can I add my dream booth to the prompt? I made a Lora and it doesn’t really look like me. Hypernetworks are sometimes iffy but sometimes amazing , so I’m here on my 2nd day of learning stable diffusion and I thank you for this tutorial. Really. I was really into RVC and so-vits-svc and I’m glad I learned all that anaconda, python and cmd prompt stuff first. Thanks again bro

[@ExploringGamez]

Hi, once I am done the training model..Model Revision and Model Epoch both values are 0..there is no change. Because of that at the end it is showing, "No executable batch size found, reached zero"

[@BrentLeBlancCG]

I think this could be a RAM error , but I’m not sure. So you have 12gb or more for your GPU? If not maybe try the Lora method. Look to those other KZbinrs I listed in description

[@aeropriest]

Thanks for this detailed tutorial, very helpful. can you pls share the link for class data images for training?

[@BrentLeBlancCG]

This is the channel I learned alot of what I did here : www.youtube.com/@Aitrepreneur Here is the video: kzbin.info/www/bejne/q6TKoqCHZbKkbck Person Training Set: github.com/aitrepreneur/REGULARIZATION-IMAGES-SD

[@zaselimgamingvideos]

i am getting this error: Exception training model: 'Refer to (link) more information on how to install xformers'. But i have xformers already installed.

[@BrentLeBlancCG]

Try to put the flag in your webui-user.bat file just find that file in the stable-diffusion-webui folder and edit in notepad to edit the file. set PYTHON="C:\Users\NAME\AppData\Local\Programs\Python\Python310\python.exe" set GIT= set VENV_DIR= set COMMANDLINE_ARGS=--xformers call webui.bat

[@NizarAnan]

Hello Brent, why dont you use the Restore faces option? From what i tested it really helps with the face.

[@BrentLeBlancCG]

Yes, you're right Restore faces does give good results. However, the issue with it when you are training on a specific person is that the face restore is using a different dataset to restore the face. I've used it before and it takes the face much further away from the likeness of the training data. If you are looking to just create images of a generic character or person and don't mind the likeness drift, restore faces can give you great results. Cheers!

[@NizarAnan]

@@BrentLeBlancCG I never thought about that! Youre right, i usually use Stable diffusion just for concept art as reference or specific idea for a full CG project. Hope artists start seeing AI as a tool and not a threat. Love your videos, Keep evolving and innovating my friend!

[@BrentLeBlancCG]

@@NizarAnan I agree, I believe AI is just another tool in our toolset that can let us do more faster. Thanks for watching!

[@BrandonFoy]

Thanks for this tutorial! Was looking for one that covered the updated Dreambooth UI and yours worked perfectly!!

[@BrentLeBlancCG]

So glad it was helpful Hauteclere! Would love to see your results :)

[@calonsultancom]

hi can you share your negatif prompts? thanks

[@BrentLeBlancCG]

(contrast:1.3), mask, hemet, (glasses:1.1), nsfw, blurry, logo, watermark, signature, cropped, out of frame, worst quality, low quality, jpeg artifacts, poorly lit, overexposed, underexposed, glitch, error, out of focus, (semi-realistic, cgi, 3d, render, sketch, cartoon, drawing, digital art, anime, manga:1.3), amateur, (poorly drawn hands, poorly drawn face:1.2), deformed iris, deformed pupils, morbid, duplicate, mutilated, extra fingers, mutated hands, poorly drawn eyes, mutation, deformed, dehydrated, bad anatomy, bad proportions, extra limbs, cloned face, disfigured, gross proportions, malformed limbs, missing arms, missing legs, extra arms, extra legs, fused fingers, too many fingers, long neck, incoherent, weird eyes, hats, weird pupil

[@calonsultancom]

@@BrentLeBlancCG thanks Brent, and thanks for this tutorial

[@BrentLeBlancCG]

@@calonsultancom No problem!

[@FinalFlameProductions]

I'm getting an error - Exception training model: 'No executable batch size found, reached zero'

[@BrentLeBlancCG]

Make sure your training photos are 512x512 and also you need a fair amount of VRAM to run it. If you don’t have much vram you can try the Lora method. Which that link for Dreambooth I have in description talks about that.

[@FinalFlameProductions]

@@BrentLeBlancCG Cheers I'll try that out.

[@azuki2919]

@@BrentLeBlancCG So I'm assuming you need more than 8gb? Even with Lora enabled im still getting that exact error

[@BrentLeBlancCG]

@@azuki2919 are all of your input images 512x512?

[@azuki2919]

@@BrentLeBlancCG Yep

[@Nutronic]

Scale Position at 07:21 doesnt seem to exist anymore.

[@Nutronic]

Where is the file outputted ?

[@a9x]

Can you share that thousand face dataset of random faces

[@BrentLeBlancCG]

Yeah sure, let me find the link

[@BrentLeBlancCG]

This is the channel I learned alot of what I did here : www.youtube.com/@Aitrepreneur Here is the video: kzbin.info/www/bejne/q6TKoqCHZbKkbck Person Training Set: github.com/aitrepreneur/REGULARIZATION-IMAGES-SD

[@a9x]

@@BrentLeBlancCG Thank you so much

[@erraddict]

I'm getting garbage when training, what gives?

[@meitoart8793]

hi settings are wrong, he forgot to to do call images per instance image

[@-CornDawg]

da fuq is this

[@BrentLeBlancCG]

This is custom AI model training using Stable Diffusion

[@BrentLeBlancCG]

Once you train this model, you can put your likeness into literally in genre