Disclaimer: Each version is different and this is not based on the arcane show's style and nowhere related to it.
V2.8 Settings:
Steps: 20+
CFG: 3+
Sampler: Whatever works for you.
Scheduler: Whatever works for you.
V1.0 Settings:
Steps: 25 - 30
CFG: 4 - 6
Sampler: Euler / Euler A / DPM++ 2M (Haven't tested with other samplers)
Scheduler: Normal / Karras (Haven't tested with other schedulers)
V1.5 Settings:
Steps: 25 - 30
CFG: 3.5 - 7
Sampler: Euler (Haven't tested with other samplers)
Scheduler: Normal (Haven't tested with other schedulers)
Note: You may have to regenerate with different seeds a couple of times it isn't 100% accurate, like many other models. You need proper pony scores. Positive Prompt should have "score_9, score_8_up, score_7_up" Negative Prompt should have "score_6_up, score_5_up, score_4_up".
V2.0 Recommended Settings:
Steps: 28 or higher
CFG: 4.5 or higher
Sampler: Euler (Haven't tested with other samplers)
Scheduler: Normal (Haven't tested with other schedulers)
Recommended Negative Prompt: worst quality, old, early, low quality, lowres, signature, artist name, signature, username, logo, bad hands, mutated hands, loli, shota, young, child
Note: You may have to regenerate with different seeds a couple of times it isn't 100% accurate, like many other models. (Optional): Proper pony scores. Positive Prompt should have "score_9, score_8_up, score_7_up" Negative Prompt should have "score_6_up, score_5_up, score_4_up".
V2.5 Recommended Settings:
Steps: 30 or higher
CFG: 6 - 9
Sampler: Euler (Haven't tested with other samplers)
Scheduler: Normal (Haven't tested with other schedulers)
Note: You may have to regenerate with different seeds a couple of times it isn't 100% accurate, like many other models. This Version is only 11% V1.0 - Refined, so proper pony scores is OPTIONAL: Positive Prompt should have "score_9, score_8_up, score_7_up" Negative Prompt should have "score_6_up, score_5_up, score_4_up". If it looks like it doesn't have a VAE make sure to regen it to a different seed to see if you can get it better looking or use my symphtia style lora version 6 with it.
Model is supported and on Tensor, CivitAI and PixAI.
Note: V1.0 Unrefined is highly unpredictable on tensor.
⚠️ IF YOU'RE GOING TO STEAL MY MODELS, MAKE IT SO THE DOWNLOAD IS ENABLED. THIS WORKS BETTER LOCALLY. - Zyrnox ⚠️
https://pixai.art/model/1819844909782505939, https://civitai.com/models/996671, https://tensor.art/models/802153041936582220
Description
Swapped to mostly the Epsilon v1.1 Clip of NoobAI and made two LoRAs for the model. Mixed in 10% clip and 21% model of Nova Furry XL. This version has more of a faded/washed out style, it's intentional VAE is baked in and it's the NoobAI V1.1 Epsilon VAE, changing the VAE will NOT change the style.
FAQ
Details
Downloads
318
Platform
SeaArt
Platform Status
Available
Created
12/1/2024
Updated
7/23/2025
Deleted
-
Files
Available On (5 platforms)
Same model published on other platforms. May have additional downloads or version variants.






