Skip to content
Research

Making video models answer in real time

Our work is about one problem: editing video while it is still being filmed, at a quality and cost that make sense for a shop window.

NoiseStep 1 · shape and colourStep 2 · detail

Technical notes

What we have written up

NOV 2026 · NOTE

Live video at 18 fps on one GPU

Chunked causal generation, a bounded attention window and two-step sampling: how Beehive keeps up with a camera.

Read the overview arrow_forward

NOV 2026 · NOTE

Scale is the hard part of jewellery

A product photo has no sense of size. A photo of the piece being worn fixes most of it; combining two pictures in one reference does not.

Ask for the note arrow_forward

NOV 2026 · NOTE

Adapters for one brand's catalogue

Training a small adapter on a brand's own videos, at the live model's two noise levels, so it keeps its streaming speed.

Ask for the note arrow_forward

Roadmap

Where Beehive goes next

  1. NOW

    Beehive Live

    Studio, store embed and extension running at 18 fps.

  2. NEXT

    Brand adapters

    Small adapters trained on a brand's own products, with permission.

  3. PLANNED

    Beehive-1

    Our own retrained model, tuned on try-on data.

  4. PLANNED

    Beehive-1 Lite

    A smaller student model for 24 fps, several shoppers per GPU and lower cost.

How we work

Measured on real sessions

We report numbers from our own deployment, say which figures are targets, and test changes side by side on the same clips before they ship.

  • Fixed test sets

    The same products and people for every comparison: right piece, right place, right colour, steady in motion.

  • Speed is a feature

    A change that improves quality but drops below live frame rates does not ship.

  • Data with consent

    We train only on data we own, have licensed, or have written permission to use.

Research questions

Can we read the full technical notes?add

Yes, on request. Contact us and tell us what you are evaluating.

Do you collaborate with research groups?add

We are open to it, especially on evaluation of try-on quality and efficient video generation.

When will Beehive-1 be available?add

It is planned. We will announce dates once training and evaluation are complete.