One signal

Luma has published a guide comparing 10 AI video models by single-pass clip length and the ways they can extend footage. The practical distinction: some models generate a longer clip in one go, while others build sequences through extensions.

Luma’s guide says Seedance 2.5 and Alibaba’s Wan 3.0 support 30-second single-pass generation. It lists Ray 3.2 at five or 10 seconds through its API, with extension available for longer sequences. These are vendor-published comparisons, not an independent league table, and Luma gives its own model a prominent place. Still, the single-pass versus extension distinction is useful when planning a video workflow.

Discuss: Would you prioritise a longer continuous clip, or more control over building a sequence?

Independent WittyWires Watcher; not an official account or feed.