models/Hexbloom - Ionsyx Revised - V0.9

Hexbloom - Ionsyx Revised - V0.9

|
11/26/2025
|
12:15:53 AM
| Discussion|
0
Portrait of a girl with fiery orange braided hair, glowing arm tattoos, black sclera yellow eyes, and sunflower tattoos surrounded by sunflowers.

Recommended Prompts

(your stuff you want goes here), masterpiece, best quality, absurdres

1girl, looking at viewer, best quality, absurdres

Recommended Negative Prompts

(your negative stuff you don't want goes here), low quality, bad, mutated hands, bad hands, artistic error, fewer digits, extra digits, extra ears

worst quality, low quality, bad, bad hands, mutated hands, artistic error

worst quality, artist sign, artist name, username, signature, patreon logo, comic

comic, worst quality, low quality, bad, artist sign, artist name, username, signature, mutated hands, logo, bad hands, four fingers, six fingers, extra digits, aged down

Recommended Parameters

samplers

Euler, Euler a

steps

20 - 25

cfg

3 - 6.5

clip skip

2

resolution

832x1216

Tips

LoRAs are very heavy in this model version; use low weights unless targeting specific character/concept/style LoRAs.

First generation images may have hand anatomy flaws, but the model attempts to minimize this.

Use style tags and LoRAs to create unique image generations.

Choose sampler/scheduler that works best for you and avoids overbaking.

For version 1.55, you need at least 'worst quality' and 'low quality' in negatives to get images.

If wanting models on Tensor or other English-language sites, contact the author.

Version Highlights

This version was made with the similar stuff in Ionsyx V1.0 - Base. This version should be better than V3.5 Ionsyx overall if you're doing your settings right.

Creator Sponsors

Quick Discord Server to Showcase Progression: https://discord.gg/UVXPdkgedh

Model available at https://civitai.com/models/1393785

⚠️ If stealing models or reposting, enable downloads to avoid bot uploads. See https://civitai.com/user/Arctenox

Disclaimer: Any version of any model may not be what you want, so try different versions until you find a version you like. Also V1.7 is the same model file as Ionsyx V5.2.


Welcome to my Frankenstein merge of Illustrious and NoobAI Checkpoint and LoRA models. All images in the thumbnails are base generations on ComfyUI or Forge. It is up to you to determine what version is best for you. By the way if you want the model on tensor let me know and what version and I'll upload it to tensor same with shakker and any other site that is in English.

Quick Discord Server that I made to Showcase Progression even Bad Anatomy images fur future and current models like this one: https://discord.gg/UVXPdkgedh

V1.7 is on the old model page as V5.2

V1.2 and V1.5 Recommended Starting Settings:

Side Note: V1.0, V1.1, V1.3, and V1.4 were skipped as these two versions are a big jump of a change. LoRAs are also very heavy in this model version so make sure to use low weights on them unless certain character/concept/style LoRAs.

Steps: 20 - 25+ (Below 20 will make colors washed it seems)

CFG: 3 - 6.5 (Usually higher than 6.5 will overbake the image since 6.5 already overbakes it a little.)

Sampler/Scheduler: What works best for you and isn't overbaked.
You do not need the below if you have an established positive and negative quality tags that works
Recommended Starting V1.2/V1.5 Pos Quality Tags:

Keep in mind some quality tags, negative and positive, effect style usually in this version the below is the basic quality tags without effecting style too much, but you don't have to use them.

(your stuff you want goes here), masterpiece, best quality, absurdres
(your negative stuff you don't want goes here), low quality, bad, mutated hands, bad hands, artistic error, fewer digits, extra digits, extra ears,
worst quality, low quality, bad, bad hands, mutated hands, artistic error,  

Note: For V1.55 you need at least worst quality, low quality to even get an image.

worst quality, artist sign, artist name, username, signature, patreon logo, comic

Note: "bad", "four fingers", and "six fingers" was trained into a lora I put into the model so it will do stuff in the negatives, specifically it was in my refiner model.

First gen may be weird with the hand anatomy as it's impossible to get rid of but I tried to get rid of the bad hand anatomy as much as I could. So there is a chance to get bad hands or good hands, or in-between.

V0.9 Recommended Starting Settings:

Steps: 14 - 25+

CFG: 4.5 - 6.5 (Usually higher than 6.5 will overbake the image)

Sampler/Scheduler: What works best for you and isn't overbaked.

As always first gen may not be what you want and remember to use style tags and LoRas if you want something unique.

This is basically Ionsyx V3.9 and higher versions, but I took the stuff that made Ionsyx V1.0 and revised it and added some more models in with V3.5 with a refiner lora for Ionsyx V3.5 that I made and is posted, also V0.9 needs tested with styles more, so feel free to test it and give some feedback if you want.

⚠️ IF YOU'RE GOING TO STEAL MY (https://civitai.com/user/Arctenox) MODELS OR POST THEM SOMEWHERE ELSE, MAKE IT SO THE DOWNLOAD IS ENABLED, IF IT'S NOT THEN I KNOW A BOT UPLOADED IT. ⚠️

https://civitai.com/models/1393785

Previous
Dark Fantasy Pulp Pinup - Illustrious
Next
Ahgot - artist style [Illustrious] - v0.1_e10

Model Details

Model type

Checkpoint

Base model

Illustrious

Model version

V0.9

Model hash

580e2bf44b

Creator

Discussion

Please log in to leave a comment.

Images by Hexbloom - Ionsyx Revised - V0.9

base model Images

illustrious Images

lora Images