AzelusLightvale/SD1.5-PandoraBox

This repository is meant to be a backup for my Stable Diffusion 1.5 models in case they ever went down on every other site I have published them on.

Note: This model is made though merging with ComfyUI and, therefore, incompatible with Model Toolkit. Do not use Model Toolkit to check/prune this model, because it will break the model.

What is this model?

ConcoctionMix (now Pandora's Box) is my experimental branch for OpenSolera, my Stable Diffusion 1.5 anime-style model. The model is my foray into ComfyUI merging and continuation of CLIP experiments (thanks @daskmaster for the CLIPs). CCM-a1 is probably one of, if not the hardest model to work with that I've made, with future models growing in complexity and pain.

The model should work the same as other Stable Diffusion 1.5 checkpoints, just much harder to merge with other models without ComfyUI. Check each version for what is used for the merge.

How to use this model?

Prompting:

This model (in my testing) uses booru tags very well, but can use natural language when necessary.

For best results, a style must be prompted (in artist and/or medium) to give this model a style to work with. Versatility is this model's strong point. Base model (without style prompt) jumps between styles very often. and can sometimes drift into 3D/2.5D art styles and sometimes realism.

Like most anime checkpoints, this one also has a female bias, but it can be countered.

This model is mainly tested against ComfyUI's prompt parser and reForge's default parser.

Character recognition should work well with really popular characters, but not much else.

Parameters:

(Anything in bold has been tested and working decently fine)

Sampler + Scheduler: Almost anything works, with recommendations being: Euler a, DPM++ 2M Karras/AYS, DPM Adaptive, UniPC-simple, DDIM (with ddim_uniform),

Steps: 20+

CFG: 6-12 (recommended range: 7-12)

CLIP Skip: 1-2

Resolution: 3:2 aspect ratio tested, 640x640, 640x960, 960x640, 512x512, 512x768, 768x512 all works to some degree

Hi.res fix: Recommended, but not necessary for close-ups. For multiple people, this is a necessity. (This is wrong as of CCM-a2. Hi.res fix for CCM-a2 is more required to get strong results)

Additional information

• Every model has ComfyUI metadata for merge information, except a3 [Mojito] due to how it was merged

Model information - a1 [Vodka]

This is the initial release of ConcoctionMix (CCM). However, it has been officially rebranded as Pandora's Box (PBX). Nothing else will change, just the naming.

Original name: OpenSolera-a6.safetensors

Due to the model's inherent differences, this model is separated from the OpenSolera branch and uploaded as its own thing.

This series of models will be named after different types of alcohol. CCM-a1's codename is Vodka, inspired by its harsh nature but ease of mixing into other drinks to "spike" them

Merge recipe

Checkpoints

OpenSolera - a5 [Serif]

LoRAs:

CluelessC's Hololive - 6.3-a10-eps-resized-512

AIOMonsterGirl - v4-LoRA/LyCORIS

Theovercomer8's Contrast Fix

DPO LoRA

CLIP

zer0int/CLIP-GmP-ViT-L-14 - ViT-L-14-TEXT-detail-improved-hiT-GmP-TE-only-HF.safetensors

VAE

PPPAniMix - v3

Model information - a2 [Vermouth]

The second version of Pandora's Box/ConcoctionMix, now with Wormwood built-in for better realism and slightly more cohesive images. More of a side-grade than an outright upgrade

Original name: CCM-a2-Vermouth.safetensors

This version's codename, Vermouth. follows upon the alcoholic drinks idea. There's been historical reports of people mixing wormwood into white wine, so this is a serendipity for the most part. I just like the name and idea.

Changelog:

Added Wormwood and CCM-a1 into the model (through trainDifference merging)

    More realism and realistic feel - partially countered by putting "realistic" and "photorealistic" into negatives.

    Lost quite a bit of color during this. This can be fixed by using a different VAE than the one bundled in this model - WD1-4-kl-f8-anime2-bless09

    Slightly better hands. In regular poses, hands are more solid and decent than before. Still fucks up in many cases

    Surprisingly decent prompt comprehension - managed to recover the "centaur" tags

    Worse area conditioning (until I can figure something out at least)

    Styles still work pretty well (if you put any of the realism stuff into the negatives)

    Both CLIP Skip 1 and 2 still works well

    Sometimes forgets details, which I don't like that much

Even more complicated merging workflow - this is getting really hard to put in 1 image lol

Merge recipe

Checkpoints

OpenSolera - a5 [Serif]

Wormwood - Alpha 1

LoRAs:

CluelessC's Hololive - 6.3-a10-eps-resized-512

AIOMonsterGirl - v4-LoRA/LyCORIS

Theovercomer8's Contrast Fix

DPO LoRA

CLIP

zer0int/CLIP-GmP-ViT-L-14 - ViT-L-14-TEXT-detail-improved-hiT-GmP-TE-only-HF.safetensors

VAE

nubby/kl-f8-anime2-blessed - WD1-4-kl-f8-anime2-bless09.safetensors

Model information - a3 [Mojito]

Starting from the SPO experiment and several other shelved merges, this model continues with the oddities found with my merge method for Pandora's Box (no, I'm not killing the CCM name fully, I just want to use both names).

Internal name: NotosMix-Complete.safetensors

NotosMix was meant to be a separate model line entirely from PBX, but I decide to keep them in the same line because of how similar they are in principle.

Mojito was chosen as the codename instead of another pure drink due to how many models got merged into this model

This model has bolder, rougher, and thicker lines, along with better knowledge in general, making it good for a first pass, while using another model (like a1 [Vodka] or OpenSolera-a5 [Serif]] for the second pass instead.

Hi.res fix is highly recommended for this model.

Merge recipe

Checkpoints

OpenSolera - a4 [Noto]

Wormwood - Alpha 1

AIOMonsterGirls - v4

AuroraONE

IdentityCrisis - EroDemonMix_0.3ws

Mistoon Anime - v3.0

True Futa - v1.5

VividOrangeMix - v1.0_Hard

Pandora's Box - v1 [Vodka]

LoRAs:

CluelessC's Hololive - 6.3-a10-eps-resized-512

AIOMonsterGirl - v4-LoRA/LyCORIS

Theovercomer8's Contrast Fix

DPO LoRA

Better half-closed eyes LyCORIS

SPO SD1.5 LoRA

Hyper-SD CFG 12-step LoRA (despite being a Hyper-SD merge, it doesn't really do that much better on steps)

CLIP

zer0int/CLIP-GmP-ViT-L-14 - ViT-L-14-TEXT-detail-improved-hiT-GmP-TE-only-HF.safetensors

VAE

nubby/kl-f8-anime2-blessed - WD1-4-kl-f8-anime2-bless09.safetensors

Model information - a4 [Last Word]

This version features a Long-CLIP text encoder model that underwent the same process as the normal CLIP-L model used in the regular model. To utilize Long-CLIP, use "Load CLIP" from ComfyUI and load the downloaded Long-CLIP model. The Long-CLIP version is still a prototype and has a lot of issues following prompts, so stick with the normal .safetensor models and only use the full package .zip if you want to test Long-CLIP for a change.

This model is also packaged in a .zip file so that everything can be included at once.

Merge recipe

Checkpoints

OpenSolera - a7 [Sake]

Wormwood - Alpha 1

OpenSolera - a5 [Serif]

LoRAs

CluelessC's Hololive LoRA - 6.3-a10-eps-resized-512

AIOMonsterGirl - v4 LoRA/LyCORIS

Honkai: Star Rail - All in One (Female) - v2.7

Monster girl concept - v06.07

Artists All In One - v1 (NovelAI 1)

Dohna Dohna STYLE one for all

Artist Style: Monster Girl / Mamono Museme - v7.3

Theovercomer8's Contrast Fix - SD1.5

DPO (Direct Preference Optimization) LoRA - SD1.5

SPO-SD-v1-5_4k-p_10ep_LoRA_webui

CLIP

Normal CLIP-L

zer0int/CLIP-Registers-Gated_MLP-ViT-L-14 - ViT-L-14-REG-TE-only-xtreme-HF-format-ckpt20.safetensors

zer0int/CLIP-SAE-ViT-L-14

zer0int/CLIP-GmP-ViT-L-14 - ViT-L-14-TEXT-detail-improved-hiT-GmP-TE-only-HF.safetensors

LongCLIP-L version

zer0int/LongCLIP-Registers-Gated_MLP-ViT-L-14

zer0int/LongCLIP-SAE-ViT-L-14

zer0int/LongCLIP-GmP-ViT-L-14 - Long-ViT-L-14-BEST-GmP-smooth-ft.safetensors

VAE

nubby/kl-f8-anime2-blessed - WD1-4-kl-f8-anime2-bless09.safetensors