The Lab · In-House Research

We develop our own film.

AI Studio began as the best way to use every frontier model. The Lab is how we go further: an in-house research group of machine learning engineers and data scientists, working from Germany, Ukraine, Canada, Italy and Türkiye, building the models and methods behind the app, trained on signal no one else has.

1M+ Creators · To Date
10+ Frontier Models in Production
03 Research Programs
Why a Lab

Every day, ten thousand active creators tell us what a good generation looks like, by what they keep, what they retry, and what they share. The Lab turns that signal into models.

Program 01
Active

Preference Learning at Scale

We train proprietary reward models on the behavioral signal of 1,000,000+ creators, systems that learn to judge images and video the way our audience does. They work ahead of the user: scoring candidates before results appear, catching weak generations for automatic retries, and learning which of the frontier models in production will serve a given prompt best, per credit spent.

The same program runs our internal evaluation harness. Every new model release is benchmarked within days of shipping, paired comparisons, confidence intervals, category-level breakdowns, so what we route to is a measurement, never a guess.

Program 02
Active

Custom Model Post-Training

Generic models produce generic results. For the tools our creators use most, photo restoration, virtual try-on, consistent characters, signature styles, we post-train our own: fine-tunes and adapters of open-weight image and video models, trained on datasets we curate in-house.

The result is a growing family of models that exist nowhere else, each tuned for exactly one job, measured against the frontier in our own harness, and shipped straight into the app.

Program 03
In Development

Efficient Inference & On-Device Generation

Speed is a creative feature. We are actively working on step distillation, quantization, and self-hosted serving, so our highest-volume tools run on our own models, faster, at a fraction of the cost, with near-instant draft previews before a full-quality render.

The long-horizon goal, planned as the Lab's multi-year flagship: models distilled compact enough to run on Apple silicon itself. Generation on the iPhone in your hand, private, instant, offline.

Now Casting · Research

The Lab is hiring.

Senior roles in machine learning, statistics, and ML systems are open now. Small team, real data, results in production within weeks.

View Open Roles › Write to the Lab

The Lab publishes inward · our results ship in the app.