Somebody has to read all of this.

The problem with demo reels — and with most reviews

Every vendor showcase is cherry-picked, and reasonably so; they're marketing. The trouble is that there's no way to compare two showreels. Different prompts, different settings, different numbers of attempts before the good one.

Independent reviews should fix that, and often don't. A great many name the tool but not the model version, so a glowing review of last year's release gets read as evidence about this year's. Some received a complimentary account. Some quote pricing that has since changed twice.

So we treat the reviews as evidence rather than as verdicts. Each source is weighted by how closely it matches the exact model version, how directly it answers the question, and how recent it is. Then we publish the finding, the confidence level and the sources behind it. The full method is here.

What we publish that others don't

Uncertainty. Almost every site in this category publishes synthesis of the same public sources we read; very few will tell you when that evidence does not support a conclusion.

So alongside each finding we publish an evidence coverage figure — how much of our criteria set the available sources actually answer. Where coverage falls below our threshold, the verdict is withheld rather than estimated. At the time of writing two of the three tools we have researched in depth have withheld verdicts, and the pages say so.

"Not established" is a real result here. It doesn't drive clicks, which is presumably why it's rare.

How this is funded

Affiliate commission. Some links here are affiliate links, and if you subscribe through one we may earn a percentage at no extra cost to you. That's the entire business model — there are no sponsored placements, no paid reviews and no pay-for-position.

Three rules keep that from bending the scores:

• No vendor accounts, briefings or review copy. We research from published sources. Where a source we cite was given a complimentary account, we note it.
• Tools that pay us nothing get researched anyway — Google Flow, Sora, Luma, Vidu and CapCut sit in the same table against the same evidence standard. They're flagged so you know which is which.
• The evidence is the evidence. Where a tool pays well and the research is thin, the research is what gets published. If that ever stops being true, this page is a lie and the site is worthless.

We also publish a running list of the tools we earn nothing from, because knowing where the money isn't is as useful as knowing where it is.

When we get it wrong

Research on fast-moving software goes stale, and sources conflict. We will be wrong sometimes. If you think a finding is off, or you know a source we missed, write to admin@testedclips.com and say why — we re-research it and publish the outcome either way, including when the re-check says the original was right.

Corrections are noted on the page rather than quietly edited in. Every page carries the date it was last researched, so you can judge for yourself how stale it is.

What this site is not

• Not a news site. We don't cover launches. We cover tools we've actually run, which is necessarily slower.
• Not comprehensive. Research is expensive in time, so we cover tools people are actually choosing between rather than everything that exists.
• Not a testing laboratory. We do not generate clips or run controlled tests, and we never imply that we do. We read the testing that exists and tell you how much weight it carries. Nothing here substitutes for trying a tool on your own work.

Who runs this

Tested Clips is run by the video editing team behind videoeditingservice.net, an online editing partner for creators, entrepreneurs, agencies and business owners who need professional video without spending their week in a timeline.

That background is why the research asks the questions it asks. We edit video for a living, so when a review says a model handles motion well, we know to ask which motion, at what length, and whether the footage survived being cut into a sequence. Those are the distinctions that decide whether a tool is usable on a paid job, and they're the ones most reviews skate past.

It also means we've seen these models fail in production rather than in a demo — which is mostly useful for knowing which claims to be sceptical of.

The disclosure that belongs here: we sell human video editing, and several tools reviewed on this site automate work we are paid to do. We'd rather say so plainly than have you find out later. If anything, it makes us harder to impress — we have no reason to overstate what these models manage, and a fairly precise sense of where they come apart. Judge it on the published clips, failures included.

Contact

admin@testedclips.com — score disputes, corrections, factual errors, or a tool you think we should be testing. We read everything; we don't accept paid placements, so please don't ask.