Research

How long does an AI real estate video actually take?

Every tool in this category says “minutes” and leaves it there. Here is the real answer, measured on 216 finished videos between 2026-07-08 and 2026-10-11, timed by the pipeline itself rather than estimated afterwards. Nothing is rounded in our favour, and it is free to cite.

The short answer

A typical listing video is finished 7 min 7 s after you press Generate. A small shoot lands nearer 3 min 12 s; a full gallery in every format can take 12 min 34 s or more. You do not wait on the page for any of it, and a videographer booked for the same listing is 2–5 days.

Render time

Where it landsTimeWhat that is
Fastest tenth3 min 12 sA handful of photos, one or two formats.
Typical (median)7 min 7 sHalf of all videos finish sooner than this.
Slowest tenth12 min 34 sLarge galleries, or every format at once.

By photo count

This is the number worth planning around. Photos are the one input you control, and render time tracks them almost linearly, because every photo is depth mapped, moved and encoded on its own.

PhotosRendersAverage time
1 to 5454 min 40 s
6 to 10707 min 2 s
11 to 15577 min 50 s
16 to 253112 min 7 s
26 or more1315 min 14 s

The narration does not grow with the gallery. A script settles at around ninety seconds however many rooms you send, so past roughly ten photos you are buying render time rather than a longer story. Eight to twelve photos is the sweet spot: enough to walk a buyer through the home, quick enough to be posted the same morning. Send the whole set if you want it, just know it is a longer wait, not a better video.

Which formats agents actually pick

Selected per render, multiple allowed. Reported because the answer surprised us: the MLS-clean export is a headline feature of every tool in this category, and one agent in ten chose it.

FormatRendersShare
YouTube18988%
Instagram Reel17782%
MLS (clean)209%
Email GIF209%
TikTok199%
WhatsApp Story199%
Facebook Ad157%
YouTube 4K-optimised10%

Method, and what these numbers are not

Time is wall clock, from the moment a job starts to the last export finishing upload. It is stamped on the job row by the pipeline itself, not reconstructed afterwards, and it includes every stage: reading the photos, writing the script, synthesising the voiceover, building the motion, encoding each format and uploading them.

It does not include queue time. Renders run one at a time, so a video submitted while another is running waits its turn first. The wait is shown on screen while it happens.

The box is a shared 4-core server that also serves five other production sites, and our own depth-motion engine runs on that server. Renders now also include a few AI-generated camera shots from an outside video model, and the time spent waiting on it is inside these figures. Dedicated hardware would be faster. These are the numbers real customers actually experienced, not a benchmark.

216 renders is a small sample from one product. The distribution is skewed by a handful of very large galleries. Treat these as indicative of this pipeline rather than of the category, and check them yourself: the timer runs on your own first video.

What comes out at the end of that wait

Frames from renders inside this dataset. Same pipeline, same measurements.

AI real estate video render time — modern white villa with a pool, from a vertical listing reel
AI real estate video render time — living room with a gallery wall in an AI-generated listing reel
AI real estate video render time — open-plan living room and kitchen in a property tour video

Frames from real PropReel renders. Every video on this page was produced by the product, not mocked up.

Cite this, or time it yourself

These figures are free to quote with a link. If you would rather check them, your first video is free and the clock is on the screen while it runs.