Wan 3.0 Text-to-Video Test: Generates 30-Second High-Quality Short Film Directly

ring_hyacinth · x · 2026-08-10

A creator shared their experience using the Wan 3.0 model to directly generate a 30-second text-to-video, completing a short film about a bar that sells strangers' memories with only subtitles added.

The author particularly praised the model's automatically generated split-screen editing techniques, calling it the strongest version in the Wan series to date. Additionally, the Wan algorithm team focuses specifically on rendering details of floors, skin, and human apple cheeks when evaluating model capabilities.

Original post →

More from Multimodal

Multimodal channel →