Ionsyx

Welcome to Ionsyx my Frankenstein of a merge model some people who used it regularly, call it and I wholly agree. Don't worry though it can guarantee from myself and others that the model is good. However I don't recommend starting with this if your new to illustrious or Danbooru in general unless you like the challenge since bad promoting will result in bad images.

Feel free to leave suggestions on what I can put in here to fix issues you may get. Plus feel free to leave critiques that are either bad or good, I like to improve.

Also feel free to publish images made from this, it can help people find the model and find your IMGs, also I like checking my model pages, so it's usually a free like from me if I like the images posted. And remember, you can always bid at minimum 1 yellow buzz on models you like at the auction page to see them get into the generator. :D

Renamed from Va'nsyx to Ionsyx so it's name is more easily remembered.

Recommended "V3 - Base" and "V3.5 - Base" Settings:

• Steps: 20+ / Depends on the Sampler and Scheduler used.

• CFG: 4.5+

• Sampler: Euler / (Mostly a personal preference and starting point.)

• Scheduler: Normal (Mostly a personal preference and starting point.)

• ADetailers / Upscalers / HiRes | Base generation for characters is kind of bad.
  Note: Similar stuff to V2.5 but you may need to use higher resolutions for decent/good base gens on full body images. Also thumbnail images are not base generation, they went through upscales and detailers.

Recommended "V2.5 - Base" Settings:

• Steps: 20+ / Depends on the Sampler and Scheduler used.

• CFG: 4.5+

• Sampler: Euler / (Mostly a personal preference and starting point.)

• Scheduler: Normal (Mostly a personal preference and starting point.)

• ADetailers / Upscalers / HiRes | Base generation for characters is kind of bad.

  
  Note: You may need to generate a few times, if you don't get what you want exactly, as always. This also has some Illustrious V2.0 in it, so people use at minimum these quality tags or other quality tags plus you can use some NL to ground the prompt:

  masterpiece, best quality, absurdres

  Also you may need to use this negative to start with or whatever other negative tags you want to use:

  worst quality, artist sign, artist name, username, signature, logo, four fingers, six fingers

  Settings V2.5 doesn't work for:

• Some negative prompt weighting. Weights above 1.4 for some negative tags can cause the image to fry.

Recommended "V2 - Base" Settings:

• Steps: 7 to 30+ / Depends on the Sampler and Scheduler used. (Probably want to use around 20 steps, if you want to upscale or whatever else)

• CFG: 4+

• Sampler: Euler Ancestral / Euler (Needs tested further)

• Scheduler: Normal (Needs tested further)
  Note: You may need to generate a few times, if you don't get what you want exactly, as always.

Recommended "V1 - Base" Settings:

• Steps: 20 - 30+ / (Probably want to use 20 steps, if you want to upscale or whatever else)

• CFG: 4+

• Sampler: Euler (Needs tested further)

• Scheduler: Normal (Needs tested further)
  Note: Sometimes a Face Detailer, Upscale, or rarely Hi-Res Fix. Also you may need to generate a few times, hands are a bit iffy.

V2 - Base Recommended Negative Prompt to Help Obtain Good Results:

worst quality, worst detail, very displeasing, artistic error, old, early, artist sign, artist name, username, signature, logo, bokeh

Note: You also can just use your own negative if it works. Also "(Needs tested further)" just means I haven't done anything past those things. You also don't really need a negative, but if you see some things wrong use one or few of those negatives I guess.

Here's a Base Positive Prompt you should use with most Illustrious + Noobai merge models:

(masterpiece), best quality, amazing quality, absurdres

Note: You don't have to weight masterpiece that's just me since I obtain results I want with it like that. ;P

Extra Unimportant Text:

Also feel free to tip, like or follow the model to show support to the model and me, only if you want because I only do this as a hobby, so the tips will go towards future LoRAs from me and bidding the model in the auction. Model will be available on tensor.art for site generation, as always (https://tensor.art/projects/847364693793431995). Files usually will be embedded with the workflow used to make it, sometimes my workflow for genning which needs optimized, so I don't suggest using it.

This model was made due to being inspired by images from princess_daphie in a DM. Thumbnail was of course still made using a 1060 Series GPU getting a 3060 12GB Upcoming Thursday (4/3/2025). - Arctenox

Note: I got the GPU a day or two early, so... the two first style versions were made on the new GPU. - Arctenox

⚠️ 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