The research behind the realism.

Six years of research, run inside Aksutra Intelligence Labs. Papers, blueprints, and custom systems that make our creatures more real than standard AI output.

4In-house systems

ZF · SPECIMEN NET

Why it matters

Realism is not luck. It is method.

When a creature looks fake, the reasons are repeatable — wrong joints, wrong weight, wrong light. Our R&D team finds those reasons, writes them down, and turns them into systems every artist here uses. That is why our output is consistent, not occasionally good.

Forge OS

One system, one standard.

Every Zayne Forge project runs on Forge OS, the studio's own operating layer. It is described at capability level — what it does for your project — on a page of its own.

Forge OSCore system

One operating layer behind every Zayne Forge production.

Forge OS is Zayne Forge's in-house operating layer for visual production — the single system behind every creature, world, reconstruction and film the studio makes. It succeeded Forge O on June 5, 2026.

Governs realism, perceptual fidelity, detailing and consistency across every output — and decides how much effort each part of a production receives.

Read the system

One experiment, start to finish

How a finding becomes a rule.

Question

Why do generated animal skins look like plastic in hard sunlight, even with high detail?

Reference

120 reference photos assembled — elephant, rhino, and camel skin in comparable sunlight, annotated for how light behaves.

Testing

48 controlled test renders, changing only three variables. Everything else locked, so the cause is provable.

Finding

Detail was never the problem. One light-setting beyond a measured threshold was — below it, every skin reads as plastic.

New rule

The threshold became a permanent check and a default inside Forge OS. The failure effectively stopped recurring.

ZF · N=48

Test batch 12 of 48 — believability vs controlled runs

From the lab

Latest findings, fresh from the bench.

Forge OS

Aug 2025

Desert-hide experiment, dissected

One hide, zero manual retouching — the full write-up of how layered anatomy references held up under desert light.

Forge OS

Jul 2025

Locomotion holds across a 100-run check

A consistent gait held across a hundred generated sequences. What worked, what still drifts, and the next benchmark.

Papersand blueprintsOur R&D team documents findings as research papers and blueprints — the basis of every system we run.

4Custom systemsBuilt in-house on customized AI models — anatomy, motion, world logic, and style consistency checks.

100%Findings appliedEvery finding that survives testing becomes a permanent pipeline rule — nothing is researched for its own sake.

Open questions

What we are working on right now.

Published honestly — the problems we have not solved yet say more about us than the ones we have.

Chasing

Wet surfaces in light

Rain, rivers, and sweat still break realism faster than anything dry. We are building a measured model of how wet skin scatters light.

Chasing

Herd-scale variation

One creature is solved; one hundred of the same species without cloning errors is not. Building variation systems that keep every individual plausible and distinct.

Chasing

Non-vertebrate bodies

Radial symmetry, exoskeletons, hydraulic motion — extending our anatomy systems to body plans evolution built very differently.

Chasing

Fossil reconstruction

Documentary-grade revival of extinct life, argued bone by bone from the published record — with every speculative choice documented for scientific review.

Start a project

Want our R&D eye on your brief?

Feasibility notes from the research team are free — and honest, even when the honest answer is no.