<h1 id="actually-played-with-it-and-found-7ouikfd9k">Actually played with it and found</h1>This model actually depicts on a 5x5 grid, and the attention is split on the 5x5 lines.Behold, the rule of 3 is broken with this model. Long live the rule of 5.This is the first Flux model I've seen with bottom-line 5x5 split attention control.<h1 id="update-v1512-is-released:-2152025-h2y44wjgl">Update - V1512 is released: 2/15/2025</h1>I'm doing an early release due to needing at least a few days to develop the captioning software to create the next version iteration; and I simply do not have the time of day.I'm releasing mostly because of the discovery with Hunyuan, and I'm also releasing a working Hunyuan merge with Simulacrum Schnell that produces very fun stuff... and it really shouldn't.This version's core features;<ul><li>does not require negative prompting</li><li>works with Flux1S models</li><li>works with Hunyuan -&gt; throws some errors just ignore them</li><li>works with the various clip_l_omega versions</li><li>unfinished, but far more robust</li><li>tons of nsfw depiction and control</li><li>absolute ton of spatial understand and entirely new methods of reasoning for the model to understand it</li><li>up to 20 identifiable characters simultaneously per image, this was not done yet so the outcomes are very hit or miss at times.</li><li>Grid and screen depiction control baked into the core.</li><li>Full comics on request, also wasn't done yet.</li><li>Tremendous amount of outfits, screen control, truncated sectioning, grid control, rotation control, offset control, size control, and even more; hit or miss due to not being completely trained yet.</li></ul>Configuration:Load using the CLIP_24_L_OMEGA for full effect, as this version of Simulacrum Schnell was heavily trained using the post SDXL CLIP_L version.I'm uncertain if this CLIP_L is the exact same version, but it's definitely pretty similar if it isn't and should work with other versions of CLIP_L omega.<h1 id="for-flux-1s-4irleomx7">For Flux 1S</h1>Positive prompt is plain English peppered with booru tags, and then use booru tags as solidifiers later on.Prompt using a mix of styles, offsets, whatever. Just talk to the thing, it'll probably understand what you want. I gave it additional intelligence for many sectors in the human world.<h1 id="whatever-you-do-don't-put-anything-into-the-prompt-that-you-don't-want-to-see.-it-is-much-more-intelligent-and-more-literal-than-it's-flux-counterpart.-c380dxjtx">Whatever you do, don't put anything into the prompt that you don't want to see. It is much more intelligent and more literal than it's Flux counterpart.</h1>This is NOT Flux how you know it. It's far more unchained and far more flexible.There's a reason why these things take so long to come out, and you'll see it here if you play with it too much.This thing IS DEFINITELY NOT a safe model, nor is it meant for all ages. It was never given a full controlled polish finetune for v2, so it will produce what you want, and you WILL get the monkey paw with it.This is NOT Schnell. This is finetuned so heavily that it barely reflects the original at times, and other times it's basically base SchnellBe VERY WARY. You WILL see monsters.<pre><code>Positive Prompt TLDR:

&lt;setting, placement, location, situation&gt;
&lt;context caption&gt;

&lt;quality&gt;&lt;styles...&gt;
&lt;subject counts&gt;

&lt;action caption&gt;

add outfits, clothing, interactions, whatever here

masterpiece, most aesthetic, very aesthetic,

&lt;superimposed captions&gt;

&lt;t5 captions for UI and overlays&gt;




Mix and match these for negative;

Negative Prompt TLDR:

sex, nsfw, explicit, questionable, safe, 
anime, 3d, realistic, 
line drawing, digital artwork, 
interpolated frame, blurry, grid_, depicted-, size_, behind, side, front, 
bad anatomy, bad hands, mutated, extra limbs, missing limbs, amputee, quadruple amputee, blood, gore, guro, 
humanoid, anthro, furry, censored, uncensored, 
lowres, good aesthetic, very displeasing, disgusting, </code></pre><pre><code>Steps: 12-58 -&gt; 32
CFG: 2.5-9 -&gt; 3.5
DCFG: 0 (flux guidance)

Samplers:
Euler -&gt; Simple
DPM-2M -&gt; Beta/Simple
DPM-2S -&gt; Beta/Simple
DEIS -&gt; SGM Uniform &lt;&lt;&lt; Just found it, so good.

Resolutions: 
1308x1308 (really close to this can't remember), very big.
1216x1216, 1216x832, 832x1216
1024x1024
1024x832, 832x1024, 832x832
768x768
512x512, 512x768, 768x512
</code></pre><h1 id="for-hunyuan-jfdopc64a">For Hunyuan</h1>It's not the easiest setup yet.This wasn't TRAINED for Hunyuan, but the next version WILL BE trained for Hunyuan. Direct Hunyuan interpolation control.However, this thing works for some reason and I still haven't figured out why.<pre><code>Steps: 12-64 -&gt;
12 works okay, but it tends to get blurry or pixelated
24 works probably the best but takes a while with many frames

Frames:
65-200
I did most of my testing around 130 or so, so don't think this is a strict rule.

CFG:
3.5 - 9
They all produce interesting outcome. 9 is actually really good sometimes.
My go-to is 6.4


LORA STRENGTH:
You will likely need two or three lora loader nodes to make it work, but it does work.
- Single Blocks 0.80
- Double Blocks 0.20
- CLIP - 1.0</code></pre>Consult the SDXL-Tag Guide for a full list of trained bbox data.Interpolative video training data used for 3d and realistic.<h1 id="update-v2-is-still-brewing-:-262025-qmcvc2nyq">Update - V2 is still brewing : 2/6/2025</h1>I've turned up the NSFW knob and broke it off. If this thing doesn't produce high grade high complexity NSFW I'll be shocked.It'll probably take another week of cooking to fully reach maturation, so bare with the time.<h1 id="update-v2-is-brewing-:1302025-sqkw9sd23">Update - V2 is brewing :1/30/2025</h1>V2 will not need a negative prompt. I've been running the same data that I ran through SDXL for an epoch to see what it'll do. It's already starting to take, and the need for a negative prompt is going away rapidly.It's about 300,000 or so images, roughly a third with plain English prompts.So either the 5x5 grid will work, or the model will burn to a crisp.It's attached to CLIP_L OMEGA V4, so be aware it'll behave a little... a little differently than you expect compared to the first version. This CLIP_L is 10 million samples smarter.<hr />I made everything so you can download it while you aren't logged in, so no login validation or keys required to auto-download it.As of V129 DEPICTION OFFSET works in a substantial way. Experimentation required.I can officially declare V122 ADVANCED PROMPT GRADE NSFW. You should be able to create the majority of common NSFW related acts and detailed situations with the current released version. I fed 1D these exact images in MULTIPLE epochs and it laughed at me.The further along the training goes, the more plain English depiction it'll be able to handle. Currently this is not a simple process, but you can work it out if you give it a bit of elbow grease.I'll be writing three articles soon based on using this model, because it's quite different than Flux1D and it's VERY VERY underestimated. The power here is substantial and responsive to training, while Flux1D often fell apart during training.<ul><li>Simple Schnell subject fixation using the rule of 3</li><li>Complex scene interactions and careful caption planning for Schnell NSFW training</li><li>Prompting NSFW interactions and adult depictions with Simulacrum Schnell V1</li></ul>I guarantee this model is far more powerful than is expected of it, and the outcomes from training are far more powerful than expected. The QUALITY is suffering a bit currently, but the additional training is showing certain traits are most definitely clearing up over time. This is only going to compound when I provide it with more training and more information for the requests.I STRONGLY advise using "Shuttle 3 Diffusion" Schnell to inference this lora. It amplifies the capabilities a large amount with less prompting. Shuttle v3.1 is okay but doesn't work as well with this lora, it's more compliant with it's own thing.Standard Flux Schnell FP16 and FP8 depict a FAIR QUALITY with the same settings now that we've hit Epoch 5/10. Many details that Shuttle 3 Diffusion is hiding or replacing with it's own training appeared from the training as emergent traits in standard Schnell, while Shuttle is still hiding the effects. FP8 is a little lesser but not by much. I actually ran the first epoch on FP8 at a higher learn rate for the baeline, so it should respond pretty well with FP8. The additional 4 were on BF16 mixed training however, which makes them substantially more powerful with the BF16 and FP16 versions of Flux Schnell. I haven't tried the BF16 version yet, but I assume it's good.Schnell FP16 requires a bit of a balancing act with prompts to make the dataset pop out, but it's not too bad. You can usually generate some fair quality stuff with a few tries and some prompt tinkering.<h3 id="be-sure-to-use-the-simv4-clip_l-no-matter-which-model-you-use-as-it's-required-for-a-proper-experience.-peut0weo3">Be sure to use the SimV4 CLIP_L no matter which model you use, as it's required for a proper experience.</h3><h3 id="you-must-use-negative-prompt-for-the-full-experience.-7vd6nzhyy">You MUST use NEGATIVE PROMPT for the full experience.</h3><pre><code>Euler -&gt; Simple
DPM2M -&gt; Simple 

Steps 28
CFG 3.5</code></pre><pre><code>V122 Epoch 5 - Generation Settings:

The model .safetensors says e4 but I mislabeled it. It's definitely e5. 

Inference:
1024x1024, 1216x832, 832x1216
1216x1216, 
1024x768, 768x1024,
768x512, 512x768, 
768x832, 832x768

rule34.xxx and rule34.us tags for 3d.
danbooru/gelbooru tags for anime.
plain English for realistic.

Nothing special REQUIRED for positive prompt, but these do help.

Positive Prompt:
anime, realistic, real, 3d \(artwork\), 3d,

&lt;CAPTION HERE&gt;

very aesthetic, aesthetic, masterpiece

#########################################
### BASE SCHNELL FP16 Negative Prompt ###
#########################################
censored, censor, bar censor, blur censor,
lowres, bad quality, low quality, bad anatomy, 
blur, depth of field, distorted, pixelated, 
bad hands, blurry hands, extra digits, missing digits, missing hands, extra hands, unexplained hands, merging, 
penis, erection, sex toy, dildo, pussy, cameltoe, 
multi penis, deformed, mutated, monster, vore,
disembodied, floating object, 
disembodied hand, disembodied foot, disembodied head,
extra feet, unexplained feet, unexplained arm, 3 legs, missing leg, missing arm, 
simple background, blurry background, cave,


####################################
### SHUTTLE BF16 Negative Prompt ###
####################################
nsfw, explicit, 
censored, censor, bar censor, blur censor,
lowres, bad quality, low quality, 
blur, depth of field, distorted, pixelated, 
monochrome, greyscale, comic, 2koma, doujin, manga,
bad hands, blurry hands, extra digits, missing digits, missing hands, extra hands, unexplained hands,
penis, erection, flaccid, pussy, cameltoe,
multi penis, deformed, mutated, monster, vore, pregnant,
cum, ejaculation, messy, unexplained white liquid,
disembodied, floating object, disembodied penis, disembodied hand, disembodied foot, disembodied head, jumping, floating, extra feet, unexplained feet, unexplained penis, unexplained arm,
simple background, blurry background, cave, </code></pre><hr />The home for the Simulacrum Schnell model zoo.The article with more detailed information about the training and process can be found <a target="_blank" rel="ugc" href="https://civarchive.com/articles/9954/schnell-v1-the-first-50k-greater-training-the-50mil-sample-flux-simulacrum-v5">here.</a>The Simulacrum Schnell versions require the Simulacrum V4 CLIP_L to function properly.The tagging template is the same as <a target="_blank" rel="ugc" href="https://civarchive.com/models/950382?modelVersionId=1064359">Simulacrum V4.</a><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/093e2fc4-d08d-4df7-89fe-ec15847028fe/width=525/093e2fc4-d08d-4df7-89fe-ec15847028fe.jpeg" />Simulacrum Schnell is protected under a slightly modified <a target="_blank" rel="ugc" href="https://www.apache.org/licenses/LICENSE-2.0.html">Apache Open Source 2.0</a> license.<pre><code>Copyright 2025 Abstract Powered

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this model or model zoo compliant component except in compliance with the License.
You may obtain a copy of the License at

 http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.

model zoo compliant component
Any code, component, image, derived image, Schnell based ai model released unto, Schnell based AI model released as derived and trained by Abstract Powered directly posted and hosted on Huggingface, Civit, or any other legal hosting service herein.

I hereby grant this direct exception to this license:
I grant the individual, small business, influencer, researcher, research facility, research groups, and small non-corporate entity direct and free of use to inference, train, replicate, modify, alter, or derive their own personal works based on all Schnell Simulacrum versions indefinitely without monetary contribution. You are free to use this model within the constraints of applicable law within your country of residence.

Special Exceptions:
Huggingface and Civit are both exempt from this rule and can profit monetarily without contribution.

Compliance:
Corporate entities, derived corporate entities, subset business entities, and for-profit research groups, or any similar group that fits the for-profit model, are to contact me directly for commercial and monetary use unless they are exempt via the exception rules.

By downloading the Simulacrum Schnell or any of it's derivatives trained and uploaded for distribution and sharing directly by Abstract Powered, you hereby accept this license.</code></pre>I'm not a lawyer. Just know that my intent is for the individual, small business, and influencer to monetarily gain from this model.Have fun everyone. I'll be posting many models.

v1512-E12 - Simulacrum Schnell Model Zoo

This is the CLIP_L. It can be directly inferenced using standard CLIP_L methods.

simv4-clip_l.safetensors

SimV4 CLIP_L

CLIP_24_L_OMEGA

v1512 [UNCHAINED]

v129 [NSFW/SFW]

v122 [SFW/NSFW]

v12 [SFW/NSFW]

v11 [SFW/NSFW]

v1 [FAIL]

<h1 id="!!!-simv4-clip_l-is-required-for-fp16-unet-!!!-edsmqgf8c">!!! SIMV4 CLIP_L IS REQUIRED FOR FP16 UNET !!!</h1><h1 id="fp16-unet-will-not-function-properly-without-the-clip_l!-jqicrbylh">FP16 UNET will NOT function properly without the CLIP_L!</h1><h3 id="f1fp16-unet-clip_l-hx2yngtb3">F1FP16 UNET 起動の為に、CLIP_Lが必要です。</h3><h3 id="f1fp16-unet-l-clip_l-5zf49777w">F1FP16 UNET 需要夹子 l CLIP_L</h3>CLIP_L was trained with 5 million samples.<h1 id="scheduled-release-11152024:-5-6:00-pm-gmt-7-k64drhhzt">Scheduled Release 11/15/2024: 5-6:00 pm gmt-7</h1>Full disclosure, I think 3.8 is probably better.<h3 id="configuration-simv4:-zx56huosg">Configuration SimV4:</h3><ul><li>The core system of Simulacrum V4 is based ENTIRELY on SUBJECT FIXATION.<ul><li>humans, humanoids, anthros, furries, robotics, machines, cars, vending machines, and anything else you can imagine or find a lora for.</li><li>A large percentage of images were trained with this tag format:<ul><li><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/093e2fc4-d08d-4df7-89fe-ec15847028fe/width=525/093e2fc4-d08d-4df7-89fe-ec15847028fe.jpeg" /></li></ul></li><li>Definitions;<ul><li>Attention Offset and Depiction tags;<ul><li>Introduced for V4, many tags were identified and their image position offsets were trained into the reaffirmation training for the core.</li><li>depiction tags are a bit wonky for now, but they work here and there, use at your own risk.</li><li>depicted-middle-left</li><li>depicted-middle-right</li><li>depicted-middle-center</li><li>depicted-upper-left</li><li>depicted-upper-center</li><li>depicted-upper-right</li><li>depicted-lower-left</li><li>depicted-lower-center</li><li>depicted-lower-right</li></ul><ul><li>full-frame &gt; 70% of the image</li><li>half-frame &gt; 40% &lt; 70% of the image</li><li>quarter-frame &gt; 25% &lt; 40% of the overall image</li><li>depicted-middle-left full-frame 1girl<ul><li>a full image depicting 1girl with her face on the center left, the system implies a bunch of details around it.</li></ul></li><li>depicted-upper-center half-frame face<ul><li>a face on the upper center, uncertain of the rest.</li></ul></li><li>depicted-lower-right quarter-frame shoes<ul><li>a pair of shows taking up about a quarter of the image in the lower right corner.</li></ul></li></ul></li><li>Captions<ul><li>Everything caption related goes first. System is trained with 255 tokens, and the captions are all less than 80 tokens, each image captioned with two complete captions for training the CLIP_L.</li></ul></li><li>Rating Tags<ul><li>safe, questionable, explicit, nsfw, sex</li></ul></li><li>Core Tags<pre><code>"b@s3_s1mul@cr7m", "anime", "3d", "3d model", '3d (artwork)', "blender (medium)", "source filmmaker (medium)", "sfm", "source film maker (medium)", "source filmmaker", "realistic", "real", "photo", "photorealistic", "illustration", "drawing", "painting", "digital", "traditional", "sketch", "render", "rendered", "model", "greyscale", "monochrome", "black and white", "color", "colour", </code></pre></li><li>Characters<ul><li>the name of characters, they often don't exist, but they can.</li></ul></li><li>Character Count<ul><li>Fairly reliable, sometimes doesn't count correctly on lower steps.</li><li>1girl, 1boy, 2girls, 2boys, a woman, two women, etc etc<ul><li>Fill in with your desired outcome.</li></ul></li></ul></li><li>Gender<ul><li>gender solidifiers like female, male, male and female, male/female, female/male, etc.</li></ul></li><li>Species<ul><li>Your human/animal etc species. Most of the scenes default to human if not specified otherwise.</li></ul></li><li>Series<ul><li>Styles your image with series/copyright tags, oftentimes completely overloads an image for now. Series/copyright tags are training for v5.</li></ul></li><li>Scene photograph type<pre><code>"photo","photorealistic","photography","photo-realistic","photo-realism","close-up","portrait","cowboy shot","dutch angle","three-quarter view","profile","headshot","full body","fullbody","half body","halfbody","close up","above view","below view","front view","rear view","side view","back view","overhead view","aerial view","aerial shot","aerial perspective","aerial photography","aerial photograph","aerial image","side shot","side perspective","front shot","front perspective","rear shot","rear perspective","back shot","back perspective",</code></pre></li><li>Substitute<ul><li>Anything the core system deems as important above other things so the system doesn't omit the tags due to the tag file having more tokens than 255.</li><li>Includes poses, size groupings, body parts, environmental interactions, and so on.<ul><li>couple, arms, legs, lying on side, hugging, kissing</li><li>midsection, navel, muscular abdomen, cleft of venus</li></ul></li></ul></li><li>Aesthetic and Quality Tags<ul><li>very aesthetic, aesthetic, displeasing, very displeasing, disgusting</li><li>highres, absurdres, lowres</li></ul></li><li>Year Tags<pre><code>"1970s", "1980s", "1985s", "1990s", "1995s", "2000s", "2005s", "2010s", "2015s", "2020s", "2025s", "1980","1981", "1982", "1983", "1984", "1985", "1986", "1987", "1988","1989", "1990", "1991", "1992", "1993", "1994", "1995", "1996","1997", "1998", "1999", "2000", "2001", "2002", "2003", "2004","2005", "2006", "2007", "2008", "2009", "2010", "2011", "2012","2013", "2014", "2015", "2016", "2017", "2018", "2019", "2020","2021", "2022", "2023", "2024", "2025",</code></pre></li><li><s>REMOVED TAGS - </s>Completely omitted from later trainings<pre><code>"tagme","bad pixiv id","bad source","bad id","bad tag","bad translation","untranslated*","translation*","larger resolution available","source request","*commentary*","video","animated","animated gif","animated webm","protected link","paid reward available","audible music","sound","60+fps","artist request","collaboration request","original","girl on top","boy on top","character request","original","original character"</code></pre></li><li></li></ul></li></ul></li><li>Use STEPS with aesthetic tags;<ul><li>10 STEPS<ul><li>disgusting<ul><li>generates the base simulacrum NovelAI V3 synthetic pose image set, often bad hands, bad eyes, bad anatomy.</li><li>great for super simple outlines and rapid pose testing.</li></ul></li><li>very displeasing<ul><li>for simplistic anime/cartoons/comics</li><li>adds a bit of detail, don't expect much with 10 steps</li></ul></li><li>displeasing<ul><li>produces some okay stuff, not bad for cartoons and simplistic anime</li></ul></li><li>aesthetic/very aesthetic<ul><li>don't, it's not enough steps</li></ul></li></ul></li><li>20 STEPS<ul><li>disgusting<ul><li>should produce and match novelai character bodies with the correct colorations and nearly correct anatomy</li></ul></li><li>displeasing/very displeasing<ul><li>will produce a surprising amount of comics, outlines, cartoons, and so on. Not a bad option</li></ul></li><li>aesthetic<ul><li>simple cartoon/anime individuals with white backgrounds</li></ul></li><li>very aesthetic<ul><li>produces low quality realism with bad anatomy, sometimes good stuff</li></ul></li></ul></li><li>30 STEPS<ul><li>disgusting<ul><li>produces fair quality novelai anime silhouettes for the system to fill in with more details when paired with the other aesthetic tags, doesn't work well on lower steps.</li></ul></li><li>aesthetic<ul><li>cartoon/anime/semi-realistic individuals with backgrounds</li></ul></li><li>very aesthetic<ul><li>produces okay anatomy but it's hit or miss when it comes to hands, feet, eyes, face, and coloration.</li></ul></li></ul></li><li>40 STEPS<ul><li>aesthetic<ul><li>sticks closer to anime, but still produces semi-realism.</li></ul></li><li>very aesthetic<ul><li>produces fair quality cartoon/anime/semi-realistic/photorealistic individuals with blurry backgrounds, mostly converts them to semi-realistic.</li></ul></li></ul></li><li>&gt;= 50 STEPS<ul><li>disgusting - anime<ul><li>produces defined NovelAI v3 bodies, hit or miss, very similar to 10 steps.</li></ul></li><li>displeasing</li><li>aesthetic</li><li>aesthetic, very aesthetic - realistic/photorealistic, realistic background<ul><li>produces the best images the system can until you up the steps. I haven't tested beyond 50 steps.</li><li>Will produce a high gradient of styles, from 3d, anime, blender, sfm, and a large multitude of weaker artist influences.</li></ul></li></ul></li></ul></li><li>USE resolutions;<ul><li>1218x832, 1338x768<ul><li>landscape, architecture, multiple characters, horizontal comics</li></ul></li><li>832x1338, 832x1216, 768x1024, 832x1024 ...<ul><li>portrait, tall images, vertical comics</li></ul></li><li>1216x1216, 1024x1024, 832x832, 768x768, 512x512<ul><li>1:1 ratio</li><li>mixed bag</li></ul></li></ul></li><li>BUILD your character from top of the image, to the bottom of the image, treat the screen like a 3x3 grid.</li></ul><h3 id="the-burndown-z14vqtbhd">The Burndown</h3><ul><li>Key differences between V38 and V4:<ul><li>Trained with 5000 HAGRID hand pose images for hand strengthening and reinforcement.<ul><li>about 80,000 samples</li><li><a target="_blank" rel="ugc" href="https://github.com/hukenovs/hagrid">https://github.com/hukenovs/hagrid</a></li></ul></li><li>Refined with 1000 of the best quality images I could find to differentiate the three core styles.<ul><li>about 50,000 samples total</li></ul></li><li>Reinforcement training based on the core images and dataset.<ul><li>about 80,000 samples</li></ul></li><li>Reinforcement coloration and body pose training.<ul><li>about 50,000 samples</li></ul></li><li>A finetuned and specific CLIP_L smarter at identifying anime, 3d, and human interactions; considerably more capable of identifying many more complex scenes and situations than before.<ul><li>5,000,000 samples<ul><li>Thank cheesechaser and wildcard. I have no idea what I taught it.</li><li>It was mostly Danbooru, Gelbooru, R34.</li></ul></li></ul></li></ul></li></ul>The definitive and powerful semi-stable model progression from Simulacrum V32. Trained with nearly 2 million samples overall from it's induction to this point, paired with it's new custom raidboss-grade 5 million sample fed CLIP_L, entirely devoted to the simulacrum core, subject fixation, pose, and relative positioning on the screen.Each lora was trained at BF16 before merging in an additive and orderly way, concatenated and merged into the core model using COMFYUI including this fully integrated burn merged lora, with carefully selected core blocks of each lora merged together in a hand picked methodology. I made multiple custom nodes and remade the checkpoint save system to specifically integrate FP scaling.Version 4 inherits more of base flux due to the added flexibility of the CLIP_L, while attaching even more behavior to the core system than ever before. High fidelity differentiation between cartoon, animation, 3d, and realistic.The CLIP_L has replaced a great deal of the added benefit from FLUX DeDistilled, providing a divergent and comparable standard F1D core to DeDistilled.Works with every lora I've tried. Enhances the tagging to produce more consistent outcomes than every other version of Consistency or Simulacrum; all at a higher fidelity in all variations, a higher context awareness, a higher training degree, a higher token count, and a higher validation system.This is the combination of a series of slow cooked, heavily tested, high image count loras combined together into the fourth edition model, dedicated heavily to individual and singular character.Multiple characters and character interactions are weaker in this version, producing a much more solid individual character within complex scenes, and more consistent base model to enhance upon.The initial lora trainings on this version show, the training can be a much higher learn rate while retaining stability (UNET LR 0.0009, CLIP_L TE 0.000001), in a much much shorter period of time, with less samples and less repeats (300 samples). This is the EXACT goal of this model in the end, is to produce any character the individual wants with little to no training, treating this UNET and CLP_L as a base model for rapid training.The preliminary testing is showing, the possibility of not needing the Simulacrum V4 model to generate images from the loras, meaning you may be able to train 30 image LORAS in a matter of 5 minutes or less on a 3080 that work on base flux, just because you trained them using the Simv4 F1D2pro unet and clip_l.

Files with SHA256: 8d436d6bdaeed00838507af25561a2e924e363d6e811d562ad0ee46891f54216