In Norse mythology, Ratatoskr (Old Norse, generally considered to mean "drill-tooth" or "bore-tooth") is a squirrel who runs up and down the world tree Yggdrasil to carry messages between the eagle perched atop Yggdrasil and to the roots of the tree to Nidhogg.This model is Fine-tuned for Animals / Creatures and Furries but can do way more :This model can do:<ul><li>Photorealistic</li><li>Digital Art</li><li>Anime</li><li>Comic</li><li>Vector Graphics</li><li>Line art</li><li>Stickers</li><li>can write short text</li><li>etc</li></ul>This is the Flux variant of <a target="_blank" rel="ugc" href="https://civarchive.com/models/192854/ratatoskr-animal-creature-and-furry">Ratatoskr </a>I fine-tuned on a merge of <a target="_blank" rel="ugc" href="https://civarchive.com/models/622876/demonflux">DemonFlux</a>Note: This on trained on a dataset at around 2500+ Pictures of the SDXL Version.So the model is in some aspects still behind the sdxl model. and still in (WIP)I recommend using the <a target="_blank" rel="ugc" href="https://civarchive.com/models/684964?modelVersionId=766628">workflow </a>with sdxl refining with the ratatoskr sdxl model<h3 id="note:-the-model-is-using-an-custom-clip_l-it's-in-the-training-and-in-the-ratatoskr-flux-workflow">Note: the model is using an custom clip_l it's in the training and in the<a target="_blank" rel="ugc" href="https://civarchive.com/models/684964?modelVersionId=766628"> Ratatoskr Flux Workflow</a></h3>You can test the workflow and model here:<a target="_blank" rel="ugc" href="https://tensor.art/template/766016514250953827">Ratatoskr - Creature - Generator [Flux Version]</a><h3 id="changelog:">Changelog:</h3>V1: Fine-tuned on 525 Pictures first versionV2:so after the first " test version" with limited dataset i added another finetune with 2000 Picturesso in total we now at 2500 pictures now so what has changed?<ul><li>more consistent</li><li>some better details even without the sdxl refiner</li><li>better with lora use than V1</li><li>changed a bit the contrast level</li><li>better with nsfw</li><li>etcto be honest this model was the most expensive i ever made so far over the time because i used external hardware again. so it will stay for now at this stage till i manage to use my own hardware.</li></ul>Note: the model has a tendency for realistic and photo to correct that use the flat colors tag ;-)V4Model got a Huge Finetune trained at around 12K images overall, also added some own loras inside the model.V4.10Added an unreleased detail Lora i made into the model include a small amount of <a target="_blank" rel="ugc" href="https://huggingface.co/spaces/ByteDance/Hyper-FLUX-8Steps-LoRA">the flux hyper lora</a><ul><li>Fixed some Color Degradation Issues caused possible by overtraining</li><li>fixed some precision issues</li><li>more stability</li><li>added some new specim and smaller details</li><li>model got a bit faster</li></ul>This Model is in Early Access because ongoing tests (so this Version is still WIP)<ul><li>More Details</li><li>more specim</li><li>also trained in some more art styles</li><li>better nsfwV5.94This model Version is Basically a new version based on V4.1 and my model <a target="_blank" rel="ugc" href="https://civarchive.com/models/897497?modelVersionId=1004290">Midgard Flux</a> + new detail loras i made especially for this model.Its unclear if this model is still a schnell model so i call it a hybrid Dev1. ;-)<ul><li>the model is more precise</li><li>more details</li><li>overall more creatures- but also a bit slower overall I can recommend this settings to start with:Sampler: DeisScheduler: BetaSteps: between 15-30 steps (25 recommended)CFG: Between 2 and 4There are 2 variants FP8 and GGUF (Q8)</li></ul>V6.67</li><li>this model Version is basically trained from zero base model was Helheim Flux.<ul><li>Updated Dataset</li><li>more detailedfor Settings:Sampler: DeisScheduler: BetaSteps: between 10- max 30 steps (16-25 recommended)CFG: Between 2 and 4</li><li>For experimental settings:This is a de- distilled hybrid so you can use till cfg 1.8 with a normal ksamplerrecommended cfg 1.2 / 20 steps / deis / kl-optimal</li></ul></li></ul><h3 id="v11-(krea)">V11 (Krea)</h3>This is basically a complete new modelits based on a specific block merge between used this method: <a target="_blank" rel="ugc" href="https://civarchive.com/articles/17766/guide-how-to-get-flux-krea-uncensored-and-improve-it-block-merging">https://civarchive.com/articles/17766/guide-how-to-get-flux-krea-uncensored-and-improve-it-block-merging</a><a target="_blank" rel="ugc" href="https://civarchive.com/models/1830404/colossusprojectkreaexperimentalbase">Colossus_Project_KREA_Experimental</a>, <a target="_blank" rel="ugc" href="https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev"> FLUX.1-Krea-dev</a><a target="_blank" rel="ugc" href="https://civarchive.com/models/1636436/sleipnir-flux">Sleipnir Flux</a>and <a target="_blank" rel="ugc" href="https://civarchive.com/models/681795/ratatoskr-animal-creature-and-furry-flux">Ratatoskr (Animal, Creature and Furry) [Flux] V6.67</a>How the model works and technical data details visit <a target="_blank" rel="ugc" href="https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev">https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev</a>after that i trained on that base a small lora and mixed in that too to get the model more stable again and for design porpoise that it looks more again like my ratatoskr model a small dataset 305 images overall) this dataset is entirely generated on my ai model output i had with my previous models and its complete synthetic made by me without any real people or celeb images.<h3 id="v11.39-(krea)">V11.39 (Krea)</h3>Readjusted and modified some block weights of V11added more parts of the <a target="_blank" rel="ugc" href="https://civarchive.com/models/681795?modelVersionId=1637227">V6.67</a> inside some other blocksadded parts of a new version of the upcoming<a target="_blank" rel="ugc" href="https://civarchive.com/models/310964?modelVersionId=2119948"> furry enhancer for Krea</a> <ul><li>model more creative</li><li>the design is closer to the original model</li><li>way better with paws / hands</li><li>better in sfw and specially nsfw</li><li>more specim inside</li><li>added better fur and details</li></ul>This model is recommended for realistic output on animals, anthro, furry and creature'sNote still in in development working currently still on some smaller details.This model is porpoise is for realistic output on animals, anthro, furry and creature's but can also used for general usagethis model is for education and scientific purposes only not for commercial usagei also can recommend using in some cases to improve hands the <a target="_blank" rel="ugc" href="https://civarchive.com/models/310964?modelVersionId=1525003">Furry Enhancer</a>An new workflow and a setting grid is in development.Made by <a target="_blank" rel="ugc" href="https://civarchive.com/user/freek22">Freek22</a><h3 id="please-check-out-my-other-models-too-and-give-like-)">Please check out my other models too and give like ;-)</h3>

Ratatoskr (Animal, Creature and Furry) [Flux Dev + Krea]

so after the first " test version" with limited dataset i added another finetune with 2000 Picturesso in total we now at 2500 pictures so what had changed:<ul><li>more consistent</li><li>some better details even without the sdxl refiner</li><li>better with lora use than V1</li><li>changed a bit the contrast level</li><li>better with nsfw</li><li>a tiny bit better in ram usage</li><li>etcto be honest this model was the most expensive i ever made so far over the time because i used external hardware again. so it will stay for now at this stage till i manage to use my own hardware.</li></ul>

ratatoskrAnimalCreatureAnd_v29.safetensors

v2.9

V11.39 Flux [Krea]

V11 Ratatoskr Flux [Krea]

V6.67 GGUF (Q5/Q8)

V6.67 FP8

v6.67 FP16

V5.94 (GGUF)(Dev1 Hybrid)

v5.94 (FP8) (Dev1 Hybrid)

v4.1 (GGUF) Q5_1

v4.1

v4.0 (FP8)

v1.0

Ratatoskr (Animal, Creature and Furry) [Flux]

In Norse mythology, Ratatoskr (Old Norse, generally considered to mean "drill-tooth" or "bore-tooth") is a squirrel who runs up and down the world tree Yggdrasil to carry messages between the eagle perched atop Yggdrasil and to the roots of the tree to Nidhogg.This model is Fine-tuned for Animals / Creatures and Furries but can do way more :This model can do:PhotorealisticDigital ArtAnimeComicVector GraphicsLine artStickerscan write short textetcThis is the Flux variant of <a target="_blank" rel="ugc" href="https://civitai.com/models/192854/ratatoskr-animal-creature-and-furry">Ratatoskr </a>I fine-tuned on a merge of <a target="_blank" rel="ugc" href="https://civitai.com/models/622876/demonflux">DemonFlux</a>Note: This on trained on a dataset at around 2500+ Pictures of the SDXL Version.So the model is is some aspects still behind the sdxl model. and still in (WIP)I recommend using the <a target="_blank" rel="ugc" href="https://civitai.com/models/684964?modelVersionId=766628">workflow </a>with sdxl refining with the ratatoskr sdxl modelNote: the model is using an custom clip_l it's in the training and in the Ratatoskr Flux WorkflowYou can test the workflow and model here:<a target="_blank" rel="ugc" href="https://**********/template/766016514250953827">Ratatoskr - Creature - Generator [Flux Version]</a>Changelog:V1: Fine-tuned on 525 Pictures first versionV2:so after the first " test version" with limited dataset i added another finetune with 2000 Picturesso in total we now at 2500 pictures now so what has changed?more consistentsome better details even without the sdxl refinerbetter with lora use than V1changed a bit the contrast levelbetter with nsfwetcto be honest this model was the most expensive i ever made so far over the time because i used external hardware again. so it will stay for now at this stage till i manage to use my own hardware.Note: the model has a tendency for realistic and photo to correct that use the flat colors tagg ;-)Made by Freek22Please check out my other models too and give like ;-)so after the first " test version" with limited dataset i added another finetune with 2000 Picturesso in total we now at 2500 pictures so what had changed:more consistentsome better details even without the sdxl refinerbetter with lora use than V1changed a bit the contrast levelbetter with nsfwa tiny bit better in ram usageetcto be honest this model was the most expensive i ever made so far over the time because i used external hardware again. so it will stay for now at this stage till i manage to use my own hardware.

ratatoskrAnimalCreature_v29.safetensors

Files with SHA256: 4e871e03a4f4d2097c3278577ba32b248813b43dc4d2edf415189ae9abc84dac