SaturdayMorning

Because this model uses Dreamlike in the merge, it carries with it a modified CreativeML license listed below.

SaturdayMorning Cartoon Merge Mix by Lotte

I sought to create a model that would do more than just anime as a cartoon style. There are a number of models on HF and CivitAI that individually do other types of animation/drawing styles, but there aren't really any that bring them together. I spent a couple weeks putting it all together.

Methodology

I observed early on that when you merge multiple models together, models deeper in the merge chain provide a kind of deep knowledge of concepts while the most recent merges provide more of a transfer of style. With that in mind, I sought out to merge in some single-item-focused models that could add conceptual flair later on. Thus, the first merges in were architecture (background), space (for backgrounds, patterns, glow, and color), jewelry (for more detailed and cohesive jewelry and accessories), and fashion (for more detailed fabrics, patterns, and fashion creativity). I merged these into the Noise Offset weights in the hopes that the final result would be offset.

After that, I started layering in a series of models that did cartoon-like styles that I liked individually. Each was done, all along the way, as an 'Add Difference'. Then, at the very end, I took the model with Noise Offset and everything else and did another 'Add Difference' merge. In a couple parts I went back and put in very small amounts (around 0.10) of the 'deep learning' models I used above. It took many hours of testing, merging, remerging, and starting all over again, but it's finally doing more than what I'd intended.

Please enjoy! Make beautiful things.

Tips

• Images can often be made directly up to 1024x768, though using high res fix settings makes better images.
• I recommend doing a high res fix at 0.65 - 0.75 strength.
• Using high res fix also helps determine scale, particularly on landscape images. The higher the resolution, the great the scale. So, if your portraits keep coming out as full body shots or vice versus, change you your orientation or starting or finishing sizes.
• My workflow is to do an initial set of 4 at 1024x768, choose one I like, then size it up through Img2Img to 1280x960, then from there I upscale.
• No idea how well this workd with artist names, I don't personally use them, but as they're part of stock 1.5, they should work.
• No idea if it produces NSFW material... probably, since at least two of the sixteen models used to make this mix do so as does stock SD1.5, but I'm not sure. I run with 'nsfw, nudity, busty' in my negative prompt to filter it out.
• You can remove nudity safely to get sexy. Move it around and/or weight it differently to experience different levels of dress. I put mine about midway through for 'sexy, but not pervy'.
• Edit your negative prompt. Heavy emphasis on terms like 'ugly', 'bad art', and 'poorly drawn' will lead to plasticky fake looking bad images and people. My negatives below.
• That said, putting (bad hands)++ and (poorly drawn hands)++ near the beginning of a negative prompt really helps with hands as a few of the models in the stack have been trained on what bad hands look like.

Models Used

• Noise Offset - https://civitai.com/models/10391/noise-offset-for-true-darkness-in-sd
• Architecture - rrustom/stable-architecture-diffusers
• Cosmos - carlosabadia/cosmos
• Chinese Jewelry Finetune - wdkwdkwdk/chinese_jewelry_fintune
• Fashion Diffusion - MohamedRashad/diffusion_fashion
• SeekArtv2 - https://civitai.com/models/1315/seekart-mega
• Pony Diffusion v3 - AstraliteHeart/pony-diffusion-v3
• Ghibli Diffusion - nitrosocke/Ghibli-Diffusion
• The_Owl_Characters_V2 - Jartemio/The_Owl_Characters_V2
• Shady Art - https://civitai.com/models/4515/shady-art-official
• Sanrio - Kafke/sanrio
• DarkFlatArtAI - https://civitai.com/models/17348/darkflatartai
• DGSpitzer Art Diffusion - DGSpitzer/DGSpitzer-Art-Diffusion
• Spop - zuleo/spop
• 2dn - LottePeisch/2dn-Diffusers
• Dreamshaper - Lykon/DreamShaper

Examples

The following examples were created using InvokeAI and the diffusers version of the model found here on HuggingFace. Negatives and prompt weights are in Invoke syntax, so if you'd like to use them you'll need to do the conversion.

Negatives

Negative Prompt for People

(bad hands)++, (3d render), amateur+, unprofessional+, (poorly drawn hands)+, text, blurry, (bad art), nsfw, busty, (poorly drawn), (bad composition)+, child, ugly, (poorly drawn eyes)+, lacklustre, beginner+, deformed, (extra digits)+, (poorly drawn face)+

Negative Prompt for Landscapes

amateur+, ugly++, unprofessional+, text, watermark, blurry+, (bad art)+, (poorly drawn)+, (bad composition)+, tiling, cluttered, lacklustre, beginner+

Cartoon Characters

Cartoon characters come out very well. Use (animated cartoon, cartooncore)++ close to the beginning or replace 'cartooncore' with some other example.

"(animated cartoon, cartooncore)++, an emo++ necromancer+ (androgynous person) with braided hair (wearing casual pants)+, baroque++, (epic composition, 8k)+, (trending on artstation hq), (in an abandoned amusement park)(portrait), (beautiful surroundings)+, celestial++, esoteric+, (dark and mysterious), cute++, (opaline highlights), pink, (pastel red), (saturated orange-), (silver) [(bad hands)++, (3d render), amateur+, unprofessional+,(poorly drawn hands)++, text, blurry, (bad art), nsfw, nudity, busty, (poorly drawn), (bad composition)+, child, ugly, (poorly drawn eyes)+, lacklustre, beginner+, deformed, (extra digits)+, (poorly drawn face)+]" -s 40 -S 1581230066 -W 960 -H 1280 -C 10 -A k_euler_a --hires -f 0.75

"(animated cartoon, cartooncore)++, a paranormal++ ghostly alchemist+ (androgynous person) with a ponytail (wearing a silk dress)+, adventurecore++, (epic composition, 8k)+, (trending on behance hd), (in an ossuary)(portrait), (beautiful surroundings)+, celestial++, esoteric+, (dark and mysterious), cute++, (fluorescent highlights), (saturated colors) [(bad hands)++, (3d render), amateur+, unprofessional+,(poorly drawn hands)++, text, blurry, (bad art), nsfw, nudity, busty, (poorly drawn), (bad composition)+, child, ugly, (poorly drawn eyes)+, lacklustre, beginner+, deformed, (extra digits)+, (poorly drawn face)+]" -s 40 -S 1581230066 -W 768 -H 1024 -C 10 -A k_euler_a --hires -f 0.75

"(animated cartoon, cartooncore)++, a (post-apocalyptic)+ male medic+ (wearing homemade clothing), kawaiipunk++, (epic composition, 8k)+, (trending on behance hd), (on an abandoned post-apocalyptic highway), (portrait), (beautiful surroundings)+, cute+, (cream highlights), (contre-jour lighting) [(bad hands)++, (3d render), amateur+, unprofessional+, (poorly drawn hands)+, text, blurry, (bad art), nsfw, busty, (poorly drawn), (bad composition)+, child, ugly, (poorly drawn eyes)+, lacklustre, beginner+, deformed, (extra digits)+, (poorly drawn face)+]" -s 40 -S 3058068838 -W 768 -H 1024 -C 10 -A k_euler_a --hires -f 0.75

"(animated cartoon, cartooncore)++, a (post-apocalyptic)+ (androgynous person) traveling trader+ (wearing casual clothing), (art deco)++, (epic composition, 8k)+, (trending on artstation hq), (in the dead remains of a forest ravaged by climate change), (portrait), (beautiful surroundings)+, cute+, (fluorescent highlights), (dappled lighting) [(bad hands)++, (3d render), amateur+, unprofessional+, (poorly drawn hands)+, text, blurry, (bad art), nsfw, busty, (poorly drawn), (bad composition)+, child, ugly, (poorly drawn eyes)+, lacklustre, beginner+, deformed, (extra digits)+, (poorly drawn face)+]" -s 40 -S 2494651887 -W 768 -H 1024 -C 10 -A k_euler_a --hires -f 0.75

Cartoon Landscapes

"(animated cartoon, cartooncore)++, a cute++ pastel+ rocket blasting off, hygge++, (epic composition, 8k)+, (trending on behance hd), (beautiful surroundings, breathtaking setting)++, (colorful, lift off, rocket)+, (electric highlights), (dappled lighting) [(bad hands)++, (3d render), amateur+, unprofessional+, (poorly drawn hands)+, text, blurry, (bad art), nsfw, busty, (poorly drawn), (bad composition)+, child, ugly, (poorly drawn eyes)+, lacklustre, beginner+, deformed, (extra digits)+, (poorly drawn face)+]" -s 40 -S 535046622 -W 1024 -H 768 -C 10 -A k_euler_a --hires -f 0.75

"(animated cartoon, cartooncore)++, a cute (home library) interior, retro++, (epic composition, 8k)+, (trending on behance hd), in a (futuristic spacestation)+, (nebula, space, beautiful view), sculptures, (beautiful surroundings, breathtaking setting)++, (fluorescent highlights), (cinematic color) [amateur+, ugly++, unprofessional+, text, watermark, blurry+, (bad art)+, (poorly drawn)+, (bad composition)+, tiling, cluttered, lacklustre, beginner+]" -s 40 -S 839576921 -W 1024 -H 768 -C 10 -A k_euler_a --hires -f 0.75

Other Formats

Additional formats also work with this model very well, despite the model not being expressly merged to create them. A couple of examples are below, and there are more in the directory.

"a vase of lily flowers on a windowsill, masterpiece++, award-winning+ Medium Closeup professional++ photograph+, (shot by a hasselblad master photographer)++, (hyperrealistic, photorealistic, insane detail)++, (bokeh, sunlight )+, ((subsurface scattering))(beautiful surroundings, breathtaking setting)++ [amateur+, ugly++, unprofessional+, text, watermark, blurry+, (bad art)+, (poorly drawn)+, (bad composition)+, tiling, cluttered, lacklustre, beginner+]" -s 40 -S 2175067135 -W 1024 -H 768 -C 10 -A k_euler_a --hires -f 0.75

"a (beautiful potion bottle) in a calm otherworldly ocean, (with a cosmic nebula inside), (digital art)+, sparklecore++, (masterpiece)++, (epic composition)++, (insane detail)++, (trending on artstation hq), (beautiful surroundings, breathtaking setting)++, (watercolor highlights)+, pink, (pale blue), (dark orange-), (silver) [amateur+, ugly++, unprofessional+, text, watermark, blurry+, (bad art)+, (poorly drawn)+, (bad composition)+, tiling, cluttered, lacklustre, beginner+]" -s 40 -S 2399755487 -W 1024 -H 768 -C 10 -A k_euler_a --hires -f 0.75

Dreamlike's License

This model is licesed under a modified CreativeML OpenRAIL-M license.

• You can't host or use the model or its derivatives on websites/apps/etc., from which you earn, will earn, or plan to earn revenue or donations. If you want to, please email us at [email protected]
• You are free to host the model card and files (Without any actual inference or finetuning) on both commercial and non-commercial websites/apps/etc. Please state the full model name (Dreamlike Diffusion 1.0) and include a link to the model card (https://huggingface.co/dreamlike-art/dreamlike-diffusion-1.0)
• You are free to host the model or its derivatives on completely non-commercial websites/apps/etc (Meaning you are not getting ANY revenue or donations). Please state the full model name (Dreamlike Diffusion 1.0) and include a link to the model card (https://huggingface.co/dreamlike-art/dreamlike-diffusion-1.0)
• You are free to use the outputs of the model or the outputs of the model's derivatives for commercial purposes in teams of 10 or less
• You can't use the model to deliberately produce nor share illegal or harmful outputs or content
• The authors claims no rights on the outputs you generate, you are free to use them and are accountable for their use which must not go against the provisions set in the license
• You may re-distribute the weights. If you do, please be aware you have to include the same use restrictions as the ones in the license and share a copy of the modified CreativeML OpenRAIL-M to all your users (please read the license entirely and carefully) Please read the full license here: https://huggingface.co/dreamlike-art/dreamlike-diffusion-1.0/blob/main/LICENSE.md