Join The Tinkerer on Whop to get early releases, private pages, and the Tinkerer Discord role - all in one place. š Join on Whop
CyberRealistic Pony Catalyst V4.0 (DMD2)
This is the special DMD2 edition of CyberRealistic Pony Catalyst, fully rebuilt for faster, ultra-realistic generations while using fewer steps.
Itās a major upgrade over the V3.0 release, delivering next-level quality for both SFW and NSFW creations.
Recommended Settings
Sampler: LCM
Schedule Type: Exponential
Sampling Steps: 8 (12 for extra fine detail)
CFG Scale: 1.5
ā” These settings are optimized for DMD2 and are not the same as my standard Pony workflow!
Patreon Exclusive Content
The full non-DMD2 version of CyberRealistic Pony Catalyst V4.0 is available exclusively on Patreon, along with other unique CyberRealistic checkpoints that are not published on Civitai.
In addition, future Catalyst models and experimental builds will also be released only on Patreon.
š Join my Patreon to access them all
Hires.fix Settings
Hires Steps: Can be the same or lower than your main sampling steps
Denoising Strength: 0.2 ā 0.5 (sweet spot for natural detail)
Hires CFG Scale: Keep it the same as your main CFG scale, or set to 1
CyberRealistic Pony Catalyst is the wild child of the CyberRealistic Pony family. Built on the foundations of CyberRealistic Pony but wired for experimentation, Catalyst isnāt here to play it safe. It's where bold ideas, unstable brilliance, and next-gen prompting chaos all collide ā and honestly? Thatās the fun of it.
This isnāt your usual polished release. Itās a testbed, a workshop, a what-if-we-just- moment captured in model form. Expect unexpected results, strange breakthroughs, and sometimes magic that might just become the next official evolution.
š§Ŗ What Makes It Different?
Built for exploration ā prompt freely, break boundaries, test weird stuff
Updated between mainline releases with no rules but vibes
Output can vary ā but when it hits, it hits hard
āļø Recommended Settings
Sampling method: DPM++ SDE Karras / DPM++ 2M Karras / Euler a
Sampling steps: 30+ Steps
Resolution: 896x1152 / 832x1216
CFG: 5
Clip Skip: 2š§ Prompting
Positive Prompt:
score_9, score_8_up, score_7_up, (SUBJECT), Negative Prompt examples:
score_6, score_5, score_4, (worst quality:1.2), (low quality:1.2), (normal quality:1.2), lowres, bad anatomy, bad hands, signature, watermarks, ugly, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, extra limb, missing limbsscore_6, score_5, score_4, simplified, abstract, unrealistic, impressionistic, low resolution, lowres, bad anatomy, bad hands, missing fingers, worst quality, low quality, normal quality, cartoon, anime, drawing, sketch, illustration, artificial, poor qualityš§ ADetailer (Optional)
Adetailer model: face_yolov9c.pt
If you only want the main face being refined set 'Mask only the top k largest' to 1.š Prompt Support
Want better outputs? Use this custom ChatGPT to auto-generate elite prompts or fine-tune your own:
š [Click here for prompts]
ā Support the Mutation
If you like living on the edge of model development, fuel the chaos with a coffee. It helps me keep making weird stuff that works.
ā ļø Keep It Responsible
Catalyst can produce sensitive content. Be mindful of what you generate. Donāt be that guy.
Description
Catalyst v1 is an experimental release that can be seen as the first beta for CyberRealistic Pony V10. Featuring improved faces, refined lighting, and other enhancements (like better Lora compatibly), it offers a glimpse into the next evolution of the series.
FAQ
Comments (14)
vpred version?
I dont know man, its very much a huge backstep.
V9 is the most realistic and has perfect poses(its less beautiful though), this is like V6 versions again.
Really? I think I might have gone too far with V9 and sacrificed some crucial features, such as LoRa compatibility. However, I see this as a solid foundation for further development. The results are significantly improved, especially in terms of the overall experience, if not necessarily in realism.
@CyberdeliaĀ What do you mean about Lora compability? Im an avid character lora user(almost 1000), they have the likeness, as in 8.5, they are less beautiful(especially with your negative), but they are working as intended.
I just tried this and get very 2.5D look again(with character loras)
@georgerobson338Ā Character LoRa is generally not an issue. It's more the Pony-specific LoRas that can be challenging. But if it works well for you, that's fantastic! As I mentioned, this is somewhat of a "experimental" checkpoint for me. I'm continually developing checkpoints with various internal branches, and this one is ideal for further development.
@CyberdeliaĀ For me its Ok. What pony specific loras dont work? I use permanent style loras too, they are working, even Illustrious and XL too. As always.
And yeah, thanx to you, even XL celebs are working perfectly.
I only miss western style one-piece swimsuits from XL(wich is default, if i prompt), Pony can only make japanese style.
This is Commander Shepard and I think this might be my new favorite Pony model on the Civitaidel.
First subjective impressions: CR 8.5, CR9a and now CR10.
CR 9a had more issues to put a character correctly into the intended scene. Much more artistic tendencies, less useful for planned scenes.
CR10 is an improvement over 9a.
Iāll give it some more runs to better compare 10 vs 8.5
This model is LEGENDARY š
Can i ask, what you find in this so exiting? We use(at least 99%) Cyber for realistic outputs. There are models, which better at realism, there are ones, which have better adherence(booru), but none of them can combine it in a good balanced way.
I find this model very semireal(similar before 6.5), and havent found any steps forward to adherence.
So can it prompt now Furry and other things or what is so new in it?
@CyberdeliaĀ I downloaded that lora, i will try it(now i have tropical, brighty background:).
You could release that photorealistic versions separately, like this.
@phillipzirin221Ā for some reason my comment is gone (?) but maybe I will release some additional Lora's
@phillipzirin221Ā I find the prompting very easy. Easier than other models. š¤·āāļø. Try the prompt "realistic textured skin" I've had a great success. As far as my comment, I just wanted to keep it simple, supportive and I do really like this model.
š¤Another observation:
v10 pushes the face of human characters a lot into the anime style (small chin, bigger eyes, etc .. such stuff). Thatās a very annoying thing if you want realistic humans.
Though: If you already have a LoRA for your character, you can suppress that tendency.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.