AI Research Lab

Generative AI.
Traditional 2D Artistry.

At ArenaX Game, we build latent control layers, automated tweening, and style-consistent vector pipelines designed to accelerate studio production while preserving the artist's signature.

Core Pillars

Our technical focus translates latent neural representations into precise, vector-native animatic curves.

Latent Control
Precision Keyframing

Direct manipulation of neural layers. Animators define critical stroke landmarks and volume limits, while the model maintains depth, style bounds, and structural consistency.

Automated Tweening
In-Between Engine

Generates style-consistent vector transitions. Calculates 2D path flows to interpolate in-betweens cleanly, preserving traditional pen speed and calligraphic thickness variations.

Vector Output
Clean SVG Curves

Our pipeline outputs fully-editable SVG vector curves instead of raster grids. Custom adapters learn your studio brush profile, keeping output scalable and resolution-independent.

Philosophy

Empowerment, Not Replacement.

We respect the craft of traditional 2D animation. AI should automate tedious bottlenecks like clean-up, coloring, and linear tweening, to give artists space to focus on acting, emotion, and storytelling. Our models translate model coordinates into artist-centric, editable vector curves.

Respect for the animator's line
100% editable vector SVGs
arenax_tween.py
import arenax as ax
adapter = ax.load_adapter("studio_ghibli_ink")
# Guided path interpolation
paths = ax.interpolate(
    keyframe_01, keyframe_02,
    steps=8, style=adapter
)
paths.export_svg("output.svg")
print(f"Exported {len(paths)} paths.")
$ python arenax_tween.py
>> Exported 284 clean bezier paths.

Partner with ArenaX

Accelerate your studio pipelines or collaborate on our generative vector research models.