Why we exist
AI video is sold on demo reels. Every launch comes with a hand-picked montage, and every platform claims the best price. What nobody shows you is the clip that came out wrong, or what one usable second actually costs once credits, plans and retries are counted.
Genativity exists to answer two plain questions with evidence you can check: which model is good at the thing you need, and where it is cheapest to run it.
The problem
- Cherry-picked results. Launch posts show the best of many attempts. Failures are left out, so two models look equally good until you try them on your own brief.
- Prices you cannot compare. The same model is resold by a dozen platforms in credits, tokens, per-second rates and subscriptions. The same 10-second clip can cost several times more in one place than another.
- Leaderboards without the evidence. A score you cannot see the clip behind is a claim, not a measurement.
What we do
How we test
- A frozen battery. Every model gets exactly the same 40 prompts. If a prompt ever has to change, it becomes a new battery version instead of silently rewriting history.
- Blind scoring. A vision model scores each clip against the original prompt without knowing which model made it, on prompt fidelity, visual quality, physics, consistency and audio.
- Checked by hand where it matters. Every score below 4 out of 5 is verified against frames pulled from the actual clip, and lip-sync lines are checked against the audio itself.
- Honest uncertainty. Leaderboards show a 95% confidence interval from each model's own runs, so two models that cannot really be separated are not presented as if they could.
The full prompt list, rubric and limits are on the methodology page.
Independence
- We pay for every clip. No model maker or platform supplies the generations we score.
- Rankings and badges are not for sale. Money never changes a score, a price or the order of a table, which is always cheapest first.
- Platforms join the radar with their agreement. We read published prices, link every one back to its source and contact a platform before adding it.
- Affiliate links are disclosed. Some links to platforms earn us a commission at no cost to you. The disclosure page explains exactly what that does and does not change.
Corrections
Automated judges are wrong in ways that sound confident. Ours has reported a sixth finger that the frames did not show; when that happens, we check the clip frame by frame and correct the score. Prices that move by more than 25% are held for a manual check before they appear, and every price we show keeps a copy of the page it was read from.
If you think a score or a price is wrong, write to the editor on the relevant page or to [email protected]. We fix errors in place and do not quietly rewrite results.
Open data
Everything behind the leaderboard and the radar is free to reuse under CC BY 4.0, with a link back to us:
- Open data: the leaderboard, every scored clip and the live prices as CSV.
- The same data on GitHub, updated daily, with a changelog of every price move.
- Benchmark badges: embeddable badges that update with each round.
- A "Cite this" box on every leaderboard, price page and study, with the citation and link ready to copy.
The team
Genativity is run by a small editorial team split across benchmarking, industry coverage and platform pricing. Each profile lists everything that person has published, so you can judge the track record directly.
Mara VossLead Benchmark Engineer
Runs the prompt battery and scores every clip before it is published. Came to AI video testing from QA automation, where she spent years building test suites designed to break things on purpose.
4 published · [email protected]
Daniel OchoaIndustry Editor
Covers model launches, shutdowns and pricing changes as they happen. Reads deprecation notices for a living so you do not have to.
31 published · [email protected]
Priya ShenoyPlatforms Editor
Tracks what AI video actually costs across the platforms that resell access to the same handful of models. Treats a pricing page as a claim, not a fact, until someone checks it.
38 published · [email protected]
Frequently asked
Is Genativity affiliated with any AI video company?
No. We are not owned by, funded by or partnered with any model maker or platform we cover. Every clip in the benchmark is a generation we paid for ourselves.
Can a company pay to rank higher or get a badge?
No. Scores and prices are never sold. Badges only exist for results a model actually earned, and they update themselves after every round.
How often do prices update?
The price radar reads every source twice a day, at 06:00 and 18:00 UTC, and keeps a copy of each page it read as evidence.
Why does a model I use score lower than I expected?
The battery is deliberately hard: it targets known failure modes like fingers, liquids and full camera orbits. Every score links to the clip it came from, so you can judge the clip yourself.
Can I use your data?
Yes. The leaderboard, every scored clip and the live prices are downloadable as CSV under CC BY 4.0. The one condition is a link back to Genativity.
Contact
- Press, partnerships and corrections: [email protected]
- Run a platform or know one we should cover? Suggest a platform.
- Want to know when a price drops? Set up a price alert, no account needed.