You can get this model, along with many others, by joining The Tinkerer on Whop. It’s one monthly membership, and you’ll also get early releases, private tools, and a bunch of extra stuff.
CyberIllustrious blends the artistic flair of Illustrious-XL with a big dose of CyberRealistic’s style. You still get all those cinematic, dreamy strokes, but now they come through with crisp, realistic detail.
If you want that painterly, elegant look but still want your images to feel sharp and lifelike, this is your model. It’s great for editorial vibes, soft glam, or even fantasy art that looks like a real photo. Basically, it just nails it.
It’s powerful, easy to use, and honestly? Prompting it is seriously fun.
📘 Get the most out of this model with the Illustrious / NoobAI Prompting: Master Guide.
⚙️ Recommended Settings
Sampling method: DPM++ SDE Karras / DPM++ 2M Karras - SGM Uniform / Euler a
VAE: is already Baked In
Sampling steps: 30+ Steps
Resolution: 896x1152 / 832x1216
CFG: 4-5About (DMD2)
The special DMD2 edition of CyberIllustrious is created for faster, cleaner generations while requiring fewer steps.
DMD2 takes the balanced improvements of the standard release and optimizes them for speed, precision, and responsiveness, making it ideal for both SFW and NSFW workflows.
Expect tighter results, smoother control, and a model that reacts beautifully to the classic Illustrious prompting style.
Recommended Settings
Sampler: LCM
Schedule Type: Exponential
Sampling Steps: 8 (12 for extra fine detail)
CFG Scale: 1.5
⚡ These settings are optimized for DMD2 and are different from my standard Illustrious workflow!
🧠 Prompting Made Easy
Positive Prompt:
masterpiece, best quality, high quality, ultra-detailed, realistic, ... rest of promptNegative Prompt examples (optional):
cartoon, illustration, anime, painting, CGI, 3D render, low quality, watermark, logo, labeldeformed, deformed face, low quality, bad quality, worst quality, (drawn, furry, illustration, cartoon, anime, comic:1.5), 3d, cgi, extra fingers, deformed teeth, watermark, logo🩺 Optional ADetailer Settings
Adetailer model: face_yolov9c.pt
If you only want the main face being refined set 'Mask only the top k largest' to 1.🔗 Resources
Backup location: HuggingFace
☕ Support the Madness
Liking what you see? Grab me a digital espresso (or a couple!). It keeps the creative energy going and helps bring more cool prompts and models your way.
⚠️ Heads Up
This model can spit out some mature stuff. Use it wisely, and please, no cursed creations—we’ve really seen enough of those.
💬 Want Better Prompts?
Use this custom ChatGPT to auto-generate polished prompts for CyberIllustrious:
🔗 [Try it now on ChatGPT]
Description
Refined this version further and keep it straightforward and adaptable.
FAQ
Comments (28)
Amazing, so far looks like a great improvement to your previous models - I've been a fan of your models since the SD 1.5 ones and while I have found some recent models lacking, looks like you hit a home run on V 2.0. Will be sharing some images shortly
Thanks. It was fun to work with V2
@Cyberdelia After playing a bit more with it, it still has some major issues with ethnicities (bollywood, desi, black, all give white girls). It does white girls really well, but really struggles with anything else (Pony and Illustrious base models I find don't have them as they were rebuilt models). Hopefully you get it fine tuned as it was one of the stronger points of the SD 1.5 models, and I think you got the realistic aspect done well with this newest 2.0 model
@Cleomoka_AI_Art It cant be prompt, like 1.5 or SDXL. You must use booru tags. "dark skinned female"
Great model! It's just a shame we're all still training on SDXL instead of 3 or 3.5. It just shows how bad they were. Makes me sad =*(.
Sorry, that was just on my mind. This truly is a great model though regardless of what's it's trained on =)
Reason why no one has trained on sd3 and sd 3.5 is due to licensing and how limited compatability they are
on AI service platforms like A1111, Forge and Comfy.
I think if someday dall-e3 can opensource, we will get more progress
Pony v7 based on AuraFlow has almost finished training, so expect a lot of people to be jumping ship on finetuning a new class of models.
sd3.5l is too resource intensive if trained with bf16, and because it has enough knowledge, if you use danbooru's tags for tagging, after training te will result in some of the tags generating images with too high a cfg look, it is more suitable to prioritize the natural language, and then complement the tags that do not have concepts in the original model, which makes the complexity rise exponentially, especially with a large amount of dataset When the dataset is large, it may require a lot of testing of which concepts are not in the original model.
@QH96 This would be awesome! I hope no one misunderstands me either. I know XL is the best we have atm & I love what Cyberdelia does for the community. I'm just over how much you have to prompt XL to get high quality results after all this time. Stable diffusion really messed up their place in the community with 3.0, Especially with models like Hunyuan video out now which is high quality video of all things. Completely uncensored & free to use. Granted my computer can't run it but still, If these AI companies don't catch up fast, they'll become irrelevant.
There's the lumina model (which allegedly is better than sd 3.5), but nobody seems to care about it.
What will be the minimum vram for pony7? Will there be Lora support? Pony team doesnt communicate shit.
@QH96 Pony v7 sounds like dead on arrival. AuraFlow is so resource intensive, building a workflow around it is already hard. Then there's the whole SDXL based ecosystem of LoRAs from Pony and Illustrious. They should have waited and trained on Flux instead.
@ElectricDreams From what I saw on Reddit, that Lumina model has crazy good prompt adherence.
Flux spoiled everything. Nobody wants to bother with a lesser model, even if that is much resource friendly. They want their hands and texts. Only out of box hardcore models can prevail. 3.5, Lumina and others have no chance.
@zachariascoburn200 yeah totally normal, tbh. It takes waaaay too many months to train a checkpoint to be "good". It's normal people don't wanna waste their time, anymore.
@zachariascoburn200 I think it's likely that it's going to be a Chinese company that releases a base model that allows NSFW out of the box
@QH96 Its high time somebody release an all rounder basic model, like stable diffusion models, but with hardcore. There are a ton anime, but nothing realistic.
@zachariascoburn200 I think the new Pony Flow V7 model, based on Auraflow, will be able to handle both anime and realistic styles. The DHXL model, released two days ago as a further training of Illustrious, seems to have multiple style options and allows you to switch between realism and anime.
@QH96 Tried with style 2(realistic). Very gorgeous, like all Duchaiten, but not realistic at all.
@zachariascoburn200 I would agree, the realism styles are a good attempt but not quite there. He probably needs more realism content in his training data. regardless of that though I think his model could serve as a good foundation for future realism finetunes. But I think it might be unlikely that his model becomes popular enough for that to happen.
@QH96 Agree. Illustrious is too strong, Pony still strong. Finetuners jumped on Illustious, than maybe Pony7.
Thanks!!!Anyway, I think the 1.5 version is the best
Amazing model thanks for sharing
Hey, can you make this available for on-site generation? I really like cyberrealistic pony, so I'd like to try this one as well.
No, unfortunately I don't have any power to decide which model is used on the on-site generator. This is decided by Civitai.
@Cyberdelia ah that's sad. Thank you for the answer anyway :)
Very good checkpoint. As promised, easy to prompt- I use a lot of Wildcards, yours work great with this checkpoint, btw. Very fast on a 10 GB 3080. Also very versatile, and realistic. Thank you for sharing.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.