Methodology
How we test listing videos, and the tools that make them.
Every comparison on this site rests on three kinds of evidence: what we saw as a customer, what each vendor publishes, and what we measured in the films. This page explains each one, what the numbers mean, and what they cannot tell you.
Updated
As a customer
We sign up and use each tool the way you would.
- 01
The same photos everywhere
Every tool gets the same 17 photos of a real five bedroom house, licensed from Pexels (see photo credits), and the same made up listing: 418 Longleaf Hollow Lane, $1,685,000, agent Jordan Avery. Nobody gets a better set to work with.
- 02
Every option, up to the card
We open every screen, toggle and template, follow every upgrade prompt to the screen before payment, and time each step. We record what the free plan can and cannot do, and every email that arrives.
- 03
What we never do
Enter a card, start a trial, post anything, connect a social account, contact sales or support, or invite anyone. Anything behind a card is described from the vendor's docs and labelled that way.
On 29 October 2026 we studied Reel-E, AutoReel and VideoGuru this way. Reel-E has no free plan, so we went through its 16 onboarding screens and 6 create steps to the card form and measured its films from customer listing websites that link the original files. AutoReel and VideoGuru let us render films on their free plans, which we saved and measured.
VideoTour.ai, Momenzo, BetterSpace and PhotoAIVideo we have not used as a customer yet. Their facts come from their public pages, help articles, terms and app store listings, and each fact box on our tested comparison says so.
From public pages
Every fact has a date and a source.
- We read pricing pages, docs, help centres, terms, app store listings and review profiles, and note the date we read each one. Comparison pages carry a "facts checked on" date and a source list.
- Numbers are copied exactly; descriptions are in our own words. We do not reuse anyone's copy, screenshots or films.
- When a vendor's own pages disagree, we say so rather than pick the number that suits us. AutoReel's October 2026 prices are an example: its pricing page, its FAQ and its app showed different figures, so we report the app's plan chooser and explain the others.
- Review counts are read on the review platform, with the date. Where a vendor offers credits for reviews, or labels its testimonials as reconstructions, we mention it next to the number.
- We recheck prices every quarter and whenever a reader or vendor tells us something changed.
In the films
What we measure, and what good looks like.
| Measure | How it is taken | What good looks like |
|---|---|---|
| Resolution, frame rate, bitrate | Read with ffprobe. Constant frame rate is checked from the packet timestamps, not the header. | 1080p for delivery; a constant rate, so motion does not stutter |
| Integrated loudness (LUFS) | EBU R128, measured over the whole film with ffmpeg's ebur128 filter. | About -14 LUFS, the level Instagram, TikTok and YouTube play at; much quieter films sound weak next to the clip before them |
| True peak (dBTP) | The highest inter-sample peak, from the same pass. | Below -1 dBTP, so the audio does not clip when an app re-encodes it |
| Point shake (px) | Up to 400 points are tracked through each shot. Each point's path minus a smooth fit over 0.43 s is its shake; we report the median, in pixels at a 1080 px short edge. | Under 0.05 px is steady; generated video often measures 0.3 px or more, which reads as shimmer or boiling |
| Camera jitter (px) | The same high pass applied to the camera's own path, built from how all the moving points change speed together. | Close to 0; a wobbling camera path looks hand held when nobody held it |
| Judder | How unevenly each point moves from frame to frame, relative to its speed. | Low; slow moves on heavily compressed files score higher, so we compare like with like |
| Peak speed and strobing | The fastest 5 percent of image motion as a share of the frame width per frame, and the share of frames above 0.5 percent. | Under 0.5 percent a frame; faster pans visibly strobe at 30 fps |
| Stalls and blend dips | Frames that stop inside a moving shot (repeats, 24 to 30 fps conversion) and frames noticeably softer than their neighbours (blended or ghosted). | Near 0 percent; repeated frames are what makes motion feel like it catches |
How the motion analyser works
It needs no reference film, so it can measure anyone's output. Cuts are found from a similarity fit between frames (RANSAC over Lucas-Kanade corners) and the motion compensated difference. Inside each shot, trimmed by 4 frames at each end, up to 400 points are tracked forward and backward with sub-pixel accuracy. A smooth camera move, eased in and out, passes through the 0.43 second Savitzky-Golay filter untouched; anything faster than about 3 Hz is shake. Numbers are frame weighted over moving shots and scaled to a 1080 pixel short edge, so 720p and 1080p files compare. Sanity checks: a synthetic ramp scores 0, and 0.3 px of white noise scores 0.28.
The editorial review
Numbers do not say whether a film is good, so we also make contact sheets (a frame every 2 seconds, and one frame per shot) and watch every film on a phone with the sound on. We note how it opens, the order of rooms, shot lengths, whether text is readable and clear of the app's buttons, how long the end cards run, whether edited or generated shots are labelled, and any invented or bent architecture.
Our own films
The same tools, pointed at us.
We run every measurement on our own demo films too, both the 1080p masters and the copies on this site, and we publish results that do not flatter us. Two examples from October 2026: our vertical cuts once panned fast enough to strobe (26 percent of moving frames over the line), which we found with this analyser and fixed with a pan speed cap; and Reel-E's frame to frame judder measures lower than ours.
| Our demo films, after the fix | Point shake | Camera jitter | Strobing frames | Loudness |
|---|---|---|---|---|
| Vertical, 1080p masters | 0.033 px | 0.014 px | 0.5 percent | -14.1 LUFS |
| Widescreen, 1080p masters | 0.028 px | 0.006 px | 0 percent | -14.1 LUFS |
The films on this site autoplay a short, low bitrate preview to keep pages fast; pressing play loads the full 1080p film, which is what we measure and what you download.
Limits
What the numbers cannot tell you.
- Taste. A steady, well mastered film can still be dull. Watch the films and judge the edit yourself.
- Every run. Generated video varies from run to run, and one listing is one sample.
- Compression. Files saved from a player or a website may be compressed harder than the original, which adds measured shake; we say which files we measured and compare like with like.
- Paid features. Anything behind a card or trial is described from the vendor's documentation, not tested.
- Speed on your connection. Render times are what we saw on the day, not a promise.
We make Property Video Generator. We try to counter that bias by testing instead of quoting, publishing our method and our own numbers, and saying on every page where the other tools are ahead. If you think we got something wrong, tell us.
Questions
Questions about how we test.
Do you take money from the tools you compare?
No. We have no affiliate links on our comparison pages, and nobody paid to be included or placed. We make Property Video Generator, which is the bias to keep in mind.
Why don't you show the competitors' films?
They are their customers' and their own work, and we do not have the right to republish them. We describe what we saw, publish the measurements, and link the vendors' pages so you can watch their examples yourself.
Why only one test listing?
Using the same 17 photos in every tool is what makes the results comparable. One listing cannot show everything a tool does, which is why we also read each tool's docs and test every option we can reach.
Can a vendor ask for a correction?
Yes. Write to us through the contact page with what is wrong and a source. We check it and correct the page, with the new date, usually within a few days.
Measure ours on your own listing.
Make a free film, download the 1080p cuts on a paid plan or the 720p ones on free, and run any test you like.