Artificial Analysis publishes AA-Video-I2V v1.0 image-to-video benchmark prompts

Artificial Analysis posts example prompts from its AA-Video-I2V v1.0 image-to-video benchmark on X. The first prompt, posted October 9, 2026, describes a deadlift personal record attempt: the lifter grips the bar, pulls slowly, locks out, screams, and drops the weight.

A second prompt, also posted October 9, describes passengers flipping newspapers open in sync while a train rocks gently, with rustling pages and train hum as audio cues.

The posts show two of the three prompts in the benchmark. The source reports do not give scoring methods, models tested, or results.