Tavus dynamic_greeting vs the first scene of a Sume video

Sume's docs describe no generated opener. You write the first line, then put it in the first scene of a video_inputs plan on POST /v1/avatar-1.0/talking-video. Tavus, by contrast, can generate one at conversation start.

Tavus facts are from its changelog; Sume facts from Generate avatar video, read 2026-10-01.

What does Tavus dynamic_greeting do?

Setting dynamic_greeting on a PAL makes conversations generate an opener from the PAL's context when there is no written greeting. You can override the field per conversation. Priority runs: a written custom_greeting, then conversation dynamic_greeting: true, then the PAL's greeting, then the PAL's dynamic_greeting.

What is the Sume equivalent of an opener?

The first entry in video_inputs. The docs describe multi-scene input as built for hooks, demos or silence beats in one composed video. A hook scene is a voice of type text with a script and a duration, plus a background prompt. The text is yours; nothing in the docs says Sume drafts it.

Can the opening be silent?

Yes. A scene with voice.type: "silence" is a non-speaking beat: duration is required and script / input_text are not allowed. Use it to hold on the avatar before the first spoken line.

What are the limits for the whole plan?

Scripts and plans are accepted when Sume estimates the video at 4-60 seconds inclusive. The docs also say current execution supports one resolved avatar per final video. For a longer piece, see avatar video longer than 60 seconds.