Meet Marsey! An adorable cat from a Telegram sticker pack. I've been trying to get SD to generate more of this character, and wanted to share my results for anyone else working on a specific 2D style.
Comparisons
a photo of a spaceman Marsey in outer space
Textual Inversion / DreamBooth
a photo of Marsey as a lifeguard
Textual Inversion / DreamBooth
a photo of Marsey as a scientist
Textual Inversion / DreamBooth
a photo of Marsey as a gardener
Textual Inversion / DreamBooth
What I've noticed:
Textual inversion:
Excels at style transfer. "elephant in the style of Marsey"
May benefit from more images. My run with 74 images performed better than the one with 3
Best results (both in terms of style transfer and character preservation) at
25,000 steps
DreamBooth:
Far, far better for my use case. The character is more editable and the composition improves. It doesn't match the art style quite as well, though.
3 images worked better than 72
works extremely well with cross-attention prompt2prompt (the "img2img alternative test" script in automatic1111's UI)
1,000 steps (30min on an A6000) is sufficient for good results
Worth mentioning - it's usable with deforum for animations
Combining the two doesn't seem to work, unfortunately. The next step might be either to directly finetune the network itself and apply one of these techniques afterwards, or possibly training the classifier.
Jump in the discussion.
No email address required.
I love this website so much.![:marseylove: :marseylove:](/e/marseylove.webp)
anyways with DreamBooth, I have gotten images that literally are almost good enough to be normal emojis on the site (haven't submitted them because I am not sure I want to open that genie's bottle). The biggest problem I have seen with DB is that a lot of time it doesn't follow the prompt very well, and there are situations in which the AI clearly doesn't know how to fulfill my request. For instance, I haven't bee able to generate a good image of marsey holding a sword - in most, the sword is hovering in front of her.
BUT the art I have been able to get with it has been really cool![:marsoyhype: :marsoyhype:](/e/marsoyhype.webp)
Jump in the discussion.
No email address required.
Jump in the discussion.
No email address required.
More options
Context
Did you try using the brackets to emphasise the prompts or something?
https://rdrama.net/post/105471/taylor-swift
Jump in the discussion.
No email address required.
are you sure you're just not telling the AI to draw them more jew-y?
Jump in the discussion.
No email address required.
More options
Context
IDK if I did with the sword prompt, but I tried with some others
Jump in the discussion.
No email address required.
img2img with prompt2prompt is the best way I've found to make Marsey editable. btw, I noticed your pruned model is based off of 2000.ckpt - if you're still using that try u1000.ckpt too (trained on three upscaled images)
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
Don't be afraid of using Img2Img to redraw a crappy mspaint of what you want
Jump in the discussion.
No email address required.
More options
Context
More options
Context