DuckyMix Anima v2 🦄🐥

Nobody Asked, Again, but I Fixed It Anyway 🔥

v1 brought the Pony vibe to Anima. v2 makes it actually look like Pony.

What changed

v1 had one annoying habit: it drifted into realism whenever it felt like it. That's beaten 🎉 I rebuilt the dataset from scratch, purely from Pony, and leaned all the way into those Pony vibes of the past. No more unprompted realism, just that Pony look I was craving, with Anima's prompt comprehension 🐴

What you get

The Pony aesthetic you remember, refined and running on Anima's architecture. Tags work, natural language works, SFW works fine too.

Style Drift Notes

You shouldn't need them anymore 😄 If something ever leans too real for your taste, the old trick still works. Add this to your negative prompt:

photorealistic, realistic, semi-realistic, 3d, photo

Recommended settings

er_sde / simple, ~30 steps, CFG 4–5, ~1MP (e.g. 832x1216)

Why

Still no good reason. The bit continues.

Notes

This is still a DuckyMix, not an OctopussMix or a Nautilus Grimoire. OctopussMix stays in its lane doing the semi-realistic anime 2.5D thing 🐙. Nautilus Grimoire does the ornate details. DuckyMix is the Pony lane, and now it actually stays in it.

Feedback welcome

Let me know if v2 nailed it or if I should've stopped at v1. I think it's great 🦄

— ducktopuss


-----------------------------------------------------------------------------------------

Nobody Asked but I Needed it 🔥

Pony Diffusion vibe merged into Anima.

That's it, that's the model. I saw a chance to bring back the Pony look with better prompt comprehension and I took it. No one asked for this. I made it anyway 😄

What you get

The Pony aesthetic you remember, running on Anima's architecture. Same saturated colors, same exaggerated proportions, but it actually follows your prompt this time 🐴

Style Drift Notes

DuckyMix can sometimes unprompted lean more realistic than expected. Here's how to steer it either way:

Want to no realism? Add to your negative prompt a combination of any of these tags:

photorealistic, realistic, semi-realistic, 3d, photo, digital artwork

Want to lean into the realism? Add this to your negative prompt instead:

anime, cel shading, flat color, 2d

Why

No good reason. Sometimes you just do the bit.

Notes

This is a DuckyMix, not an OctopussMix. OctopussMix stays in its lane doing the semi-realistic anime 2.5D thing 🐙. DuckyMix is the other lane.

Natural language prompting works fine, tags work too.

Feedback welcome

Let me know if this was a mistake or a genuinely good idea, I think its great 🦄

— ducktopuss