Generate with AI Content Drop →
AI Video Arena

Data & snapshots

Last verified/updated: 2026-10-06

Every fact on this site is generated from a data file, and the data file is published. The downloads below contain the current catalogue — per model, per category: vendor, licence, release, maximum resolution, maximum duration, native audio, access routes and the vendor URL each fact was read from. If a figure on a page ever disagrees with the snapshot, the snapshot is right and the page is stale; that is a bug and belongs on the corrections page.

Two formats are provided. The CSV is the stable, diff-friendly contract: one row per model per category, columns in a fixed order, safe to import into a spreadsheet or a data pipeline. The JSON carries the same rows plus snapshot metadata — generated date, arena status and the minimum-vote threshold — for machines that want to verify provenance. Both files are regenerated on every rebuild, so a repeated download of the same snapshot date yields byte-identical content.

Both files carry the rating columns the live snapshot will use, filled in the only honest way available today: status is prelaunch, votes is 0 and score is empty for every row. Integrators can build against the real schema now; when voting starts the same columns start carrying numbers and nothing else about the format changes.

Licensing of the data: the snapshot is factual tabular data about published model specifications and is released for reuse with attribution — a link back to this page is enough. Each row also carries the vendor URL it was read from, so the primary source travels with the data. Clips shown in future battles are not part of the snapshot and will be governed by their own per-clip terms.

Downloads