30-second video with sound: Pika Video Studio vs Sume ids

Pika's new Video Studio makes a short clip or a multi-shot video up to 30 seconds long, with sound. On Sume, seedance-2.5 accepts 4 to 30 seconds and wan-3.0 accepts 2 to 30 seconds. The Genjutsu motion-transfer model also reaches 30. Every other catalog model is capped at 15 seconds.

Sume's limits are from Video Router, read 2026-10-01; Pika's claim is from its September 17 post.

Which Sume models reach 30 seconds?

The docs state limits per model, and the models endpoint reports each model's supported durations, so check it before you set duration.

Does the clip come with sound?

generate_audio asks for audio alongside the video and defaults to the model's own audio capability. Whether a given model has audio is reported as generate_audio in its GET /v1/videos/models entry. Do not assume it from the id.

Is there a multi-shot option?

The Sume video pages describe a single prompt producing a single clip and do not describe a multi-shot mode. To build a multi-shot piece, generate each shot as its own request and join them with Timeline 1.0. Steps are in make a 1-minute video and extend a video past 30 seconds.

What happens if I ask for more than a model allows?

Pick the id by the length you need, not the other way around. For a 20-second ask, seedance-2.5 and wan-3.0 fit; a model capped at 15 does not. The 2-second floor of wan-3.0 versus the 4-second floor of seedance-2.5 matters for short beats too; see the Wan 3.0 scene-length guide.