I created this embedding for SD 2.x 768x768 models, it turns everything into your favorite Christmas classic AniMagic stop motion style as popularized by Rudolf the Red Nosed Reindeer and Santa Claus is Coming to Town among several others produced by the same studio!
The Unreleased Christmas Stop Motion Mario Kart Movie!
Prompt: mario kart toy, (rnknbss16 :1.3), highly textured, figurine
Negative prompt: cgi, 3d render, videogame
Steps: 34, Sampler: Euler a, CFG scale: 7, Seed: 2737353293, Size: 768x768, Model: SD 2.0_Standard_v2-1_768-ema-pruned, Denoising strength: 0.79, Mask blur: 3, aesthetic_score: 4.9
The Upcoming Stop Action Pikachu Movie!
Prompt: pikachu in the style of rnknbss16
Steps: 30, Sampler: Euler a, CFG scale: 7, Seed: 459369051, Size: 768x768, Model: SD 2.0_Standard_v2-1_768-ema-pruned, aesthetic_score: 5.2
Prompt: pikachu in the style of rnknbss16-100
Steps: 30, Sampler: Euler a, CFG scale: 7, Seed: 4076512951, Size: 768x768, Model: SD 2.0_Standard_v2-1_768-ema-pruned, aesthetic_score: 5.2
Some 2022 Holiday Ads for the Latest Celebs!
Donald Trump
Prompt: a close up of (donald trump:1.) in the style of (rnknbss16 :1.0)
Negative prompt: blurry, text, words
Steps: 29, Sampler: Euler a, CFG scale: 7, Seed: 1397465632, Size: 768x768, Model: SD 2.0_Standard_v2-1_768-ema-pruned, aesthetic_score: 5.4
Morgan Freeman
Prompt: morgan freeman in the style of (rnknbss16 :1.0)
Steps: 29, Sampler: DPM++ 2S a Karras, CFG scale: 7, Seed: 1868403973, Size: 768x768, Model: SD 2.0_Standard_v2-1_768-ema-pruned, aesthetic_score: 5.7
Barack Obama
Prompt: barack obama in the style of rnknbss16v2-775
Steps: 47, Sampler: Euler a, CFG scale: 7, Seed: 3661737292, Size: 768x768, Model: SD 2.0_Standard_v2-1_768-ema-pruned
And Lastly, The Remake of A Lifetime, Hogwarts Castle From the New Harry Potter Series
Prompt: Hogwarts school of witchcraft and wizardry in the style of (rnknbss16 :1.0), highly detailed, intricate
Negative prompt: blurry
Steps: 60, Sampler: Euler a, CFG scale: 7, Seed: 2909664084, Size: 768x768, Model: SD 2.0_Standard_512-depth-ema, Denoising strength: 0.66, Mask blur: 3, aesthetic_score: 6.2
Notes on the use of these:
So I didn't really get a chance to fine-tune them as well as I would have liked, but I wanted to get them out there for people to enjoy so I've included the best of what I have.
All of these were trained with 90-ish upscaled screen grabs from high quality DVDs of just the 2 movies mentioned above. I did use some of the letters, and postcards, and packages from the opening credits scenes in hopes to be able to reproduce those or something similar (haven't tried) so you will probably want to include the usual "words, text, letters, logos, watermarks..." in your negative prompts to try to weed those out. I also included some of the limited 2d artwork found in those movies, again in hopes to be able to generate that style as well. but that hasn't seemed to affect much except possibly when generating things that have a lot of 2d variations (i.e. comic book characters) so specifying 3d, or that you want a doll of the thing, or a model, or toy of the thing might help a lot with prompting. Otherwise, just saying " thing in the style of rnknbss16" should do the trick!
The Models:
They're all 16 vectors.
rnknbss16: pretty good but was trained too far and/or fast and tends to make hybrid elf/Santa creatures out of everything and is hard to get it to do anything else, although if your concept is strong or present enough in the model it can do pretty well (i.e. Cinderella's castle which is on EVERYTHING Disney).
Models rnknbss16-100 through rnknbss16-150 do much better, however these do less well with people and faces, they're better suited for things, creatures, animals, scenery, places, etc.
rnknbss16v2: pretty sure this one is overtrained by a good deal, but you might have success with it.
rnknbss16v2-750 and rnknbss16v2-775 are the sweet spot for people and characters with this v2 model, it also tends to get clearer outputs without looking as "fuzzy" or "blurry" and almost as a similar quality as VintageHelper embedding.
Which brings me to mixing this with things:
Using VingateHelper tends to enhance the "old school" vibes and film grain look as well as thematic props and other elements that may appear in the scene, and PhotoHelper embedding tends to create more "clay" models out of things, like with the Hogwarts castle it made it a wide angle clay diorama model of sorts which was cool and unexpected (see below).
Prompt: Hogwarts castle in the style of (rnknbss16 :1.2), highly detailed, very textured, intricate, shallow depth of field, photohelper
Negative prompt: blurry, text, words
Steps: 50, Sampler: Euler a, CFG scale: 7, Seed: 3448665914, Size: 768x768, Model: SD 2.0_Standard_v2-1_768-ema-pruned, aesthetic_score: 5.6