
Short-term rental and commercial property marketing
The pipeline that turns a property into a finished marketing video
Rental and commercial listings need video, and video is the one asset that refuses to scale. So we built the system that produces it end to end: footage in, a cut, graded, scored, captioned marketing video out, with no editor in the middle. The film playing behind this page came out of it.
- 0
- Editors in the pipeline
- 1
- Upload, one finished video
- 24/7
- Rendering unattended
- 4
- Aspect ratios per run
The challenge
What the old system could not do
A short-term rental books better with video, and a commercial space leases faster when a prospect can walk it before they drive out. Everyone in the market knows this. Almost nobody does it at portfolio scale, because one property video means a videographer on site, an editor for two days, and an invoice that only makes sense on a trophy listing.
So video stays rationed. An operator with forty units cannot shoot forty videos, which means thirty nine listings go out with stills while the whole marketing budget concentrates on the one property that photographs best. The units that most need help selling themselves are exactly the ones that never get the asset that would do it.
None of that work is actually bespoke. It is the same shot order, the same pacing, the same music bed, the same caption style, over and over, with only the property changing. Repetitive skilled work with a fixed structure is the thing a pipeline is for, and it had not been built because the pieces only became good enough recently.
What we built
Property Marketing Video Engine
- 01
Whatever footage the property already has
No shoot is commissioned. The pipeline ingests listing photos, a phone walkthrough, drone passes, a floor plan, in whatever mix exists, then normalizes resolution, frame rate, color, and orientation so a mismatched pile of sources becomes one coherent set of clips before any edit decision is made.
- 02
Shot selection and pacing decided by model
Vision models score every clip and still for exposure, stability, composition, and which room it shows, then assemble a shot list that opens on the exterior, moves through living space, kitchen, and bedrooms in the order prospects actually care about, and closes on the view. Cuts are placed against the beat of the chosen music bed rather than on a fixed interval.
- 03
Script, voice, and captions written to the listing
Property details drive a script written to that specific property rather than filled into a template, read in a synthetic voice, and burned in as captions timed to the read. The finished cut renders in every aspect ratio the channels want, so one property yields a landscape listing video and vertical social cuts out of a single run.
- 04
One approval, then it goes out
A finished video lands in a review queue with its script and shot list beside it. An operator approves it or sends it back with a note in plain language, and the pipeline re-renders against that note instead of returning to a human editor. That gate is the only human step in the system, and it takes about a minute.
The outcome
What changed once it shipped
- A property goes from raw footage to a finished, captioned, scored marketing video without an editor ever opening a timeline.
- Video stops being reserved for the trophy listing, because the fortieth video costs what the first one did.
- One run produces the landscape listing video and the vertical social cuts together, so no channel waits on a second pass.
- The only human step is a one minute approval, and a rejection is a note rather than a reshoot.
- The pipeline renders unattended, so a unit added on a Friday has its video by Saturday.
Built with
The stack underneath
- TypeScript end to end
- Vision models for shot selection
- Generative script and voice
- GPU render pipeline
- Queue-backed job orchestration
- Managed object storage
Ready for the system your business runs on?
We start by mapping how the operation actually works, then build the platform that fits it. If your industry is still open, the first conversation costs an hour.