Image models
Storyboards, background plates, color script. Generated per shot from the shotlist; cut and refined by a director.
An AI-native animation studio. A small team using image and video models to build anime, from frames to full episodes.
Aventos is in build. We're a few people putting image and video models inside a real animation pipeline. The tools draw, in-between, color and composite — humans direct, taste, and pick every frame that ships.
Storyboards, background plates, color script. Generated per shot from the shotlist; cut and refined by a director.
Character-conditioned generation per cut, driven by the board. Action passes re-rolled and curated; quiet shots usually land first take.
Depth, grain, FX, light leaks, color. This is where generated frames stop looking generated and become a finished cut.
Score, ADR, foley, mix. The episode leaves the studio at broadcast quality, ready for the window it'll ship to.
Range matters more than any one cut. Quiet rooms to fast action, intimate to spectacle — the same pipeline, the same hands.
Six stages. Models do the heavy lifting between human passes.
Adaptation pass, beat sheet, shotlist. Pure human work — the spine of the episode is written before any frame is generated.

Rough boards generated per shot from the shotlist, then a director cuts and re-times. The board is the contract for every later stage.

Background plates and color script. Lighting decisions get locked here — every later generation references the plate, so the look stays coherent across the episode.

Character-conditioned video generation per shot, driven by the board. Action passes get re-rolled and curated; quiet shots usually land on the first take.

Comp the shot — depth, grain, FX, light leaks — and grade. This is where AI-generated frames stop looking generated.

Sound mix, subs, deliverables. The episode leaves the studio at broadcast quality, ready for whichever window we ship to.
