
01
Drift — Generative Film
A generative film system for a fashion week opener, trained on 40,000 frames of fabric motion and rendered live on-site.

Building moving images with code and machine learning — generative systems, AI film, and real-time visuals for brands who want to feel ahead.

How it works
Every model trained from scratch. No reused generic checkpoints, no off-the-shelf "AI look."

Understand the brief, the brand world, and what "never seen before" means for this client.

Build or fine-tune the model on the right dataset — this is where the look gets found.

Run generations, curate ruthlessly, then hand off final renders or a live generative build.

I train my own models, break other people's, and turn both into images nobody has seen before — for fashion houses, labels, and studios who want their visuals to feel five minutes ahead. No stock AI look: every piece is trained, tuned, and art-directed by hand.

Custom-built systems that keep producing new, on-brand visuals long after the first delivery.
Moving images generated frame by frame, not edited together from stock footage.
Every model trained on your own references — never a generic, off-the-shelf checkpoint.
Curating and shaping the output until it has a point of view, not just volume.

Seven years training and directing generative systems.
40,000+ frames used to train a single film system (Drift).


3 industries — fashion, music, and installation art — run on the same process.

Seven years training and directing generative systems.

40,000+ frames used to train a single film system (Drift).

3 industries — fashion, music, and installation art — run on the same process.
Testimonials
A few of the studios and labels who've had Nova build something for them.
“Nova doesn't just generate images — she directs them. The work looked like nothing else in the show.”
Elena Marsh
Creative Director, Static Season
“We asked for something that felt impossible. She trained a model from scratch and delivered exactly that, on time.”
Darius Okafor
Founder, Nocturne
“Technically sharp and genuinely artistic — a rare combination in this space. Our album art has never gotten this reaction.”
Sasha Lind
A&R, Field Records
FAQ
If it's not here, it's one email away.
Every project gets its own trained or fine-tuned model. Reused generic models are how you end up with the "AI look" — that's exactly what I avoid.
Both. Installations and events often run real-time systems; campaigns and brand visuals are usually curated and rendered for final delivery.
Live input — audience sound, sensor feeds, an API, whatever the piece needs — gets piped straight into the generative engine, so the visual reacts as data comes in instead of looping a pre-rendered clip. That's the same pipeline behind Echo Chamber: it responds, it doesn't just replay.
A creative brief, any brand or reference assets you want the system trained on, and a sense of what "never seen before" means for you.
Yes — most clients haven't. Part of the job is translating the technical process into something a non-technical team can approve with confidence.
Contact Us
Currently open for new commissions from October 2026.





