The problem with catalogue video

A retail catalogue is not a campaign. It is thousands of SKUs, each needing the same treatment, to the same standard, on a schedule that never stops. Traditional production prices this per shoot day, which means video is reserved for hero products and the long tail of the catalogue never gets any at all.

Rubick came to us with the version of this problem that breaks conventional production entirely: apparel SKUs supplied as flat product photography, needing on-model video, at a volume no studio could shoot.

What we built

The input was a product image. The output was a finished vertical video of that garment worn by a model, in an environment appropriate to the category. No shoot, no fitting, no model booking, no reshoot when a SKU changed.

Each item moved through the same pipeline: model insertion for products supplied without one, face treatment for consistency across a batch, and a background chosen to suit the product category and where the video would end up being used.

Why volume needs a QC process, not just a generator

Generating a video is the easy part. Generating fourteen hundred of them that a client will approve is a different discipline, and it is where most AI production falls over.

We ran delivery in batches of roughly 500, tracked per SKU in a shared sheet, with every item carrying its own status. Client feedback came back against specific SKU codes rather than as general notes, which meant a rejection was actionable rather than a discussion. Revisions moved through a separate overlay pass and returned into the same tracking.

Over 1,400 videos were approved and delivered this way, with a further revision round of more than 250 handled inside the same process.

What this is actually useful for

Any brand whose catalogue is larger than its production budget. Marketplace listings that convert better with video but cannot justify a shoot per SKU. Seasonal ranges that turn over faster than a production calendar can accommodate.

The unlock is not that AI makes video cheaper. It is that throughput and quality control stop being in tension, so the long tail of a catalogue becomes reachable for the first time.