How We Test

PickaSi tests AI video tools by actually running them.

• Every tool we rank has been run by us on the same fixed briefs and the same fixed assets, under our published test protocol. Same prompt, same script, same source files — for every tool in a comparison.
• For every run we log: tool and version, plan and price paid, credits before and after (including retries), wall-clock time, number of generations and failures, export specs, and watermark status. We calculate cost per usable finished minute from those logs, retries included.
• Scores use published weights: output quality for the job 30%, cost per usable finished minute 25%, speed and reliability 15%, languages (EN/ES/ZH) 10%, rights and terms clarity 10%, editing control 10%.
• Raw outputs and receipts are archived with each test. If we say a tool did something, there is a log and a file behind it.
• Anything we have not run ourselves is labeled “vendor claim — not tested.” Vendor copy is never presented as our review.
• Prices carry a “last verified” date. Verdicts are reviewed at least every 90 days; a price change triggers an update the same week.
• Last updated: October 29, 2026