Snapchat short dramas: a 30 s to 5 min episode budget

Snap and Ipsos describe short dramas as vertical, episodic videos that run from 30 seconds to five minutes. On Sume you build an episode of that length as several generated shots joined by Timeline 1.0, which renders a 1080×1920 MP4 by default. The shot count follows from the clip length your chosen model accepts.

The format description is from Snapchat's research post; limits are from Sume's Timeline and video generation docs, read 2026-10-01.

How many shots does an episode need?

Divide the episode length by the clip length. The docs list seedance-2.5 at 4–30 seconds per request and Gemini Omni Flash 1.1 at 3–10 seconds in 16:9 or 9:16, so the same five-minute episode is a different number of shots depending on the model.

How do I join the shots into one vertical episode?

Timeline takes ordered video[] slots plus an audio spine (audio.duration_seconds, 1–1800 s, or audio.mode: "silence"). Slots after the first can carry a transition of fade, wipeleft, wiperight, slideup, slidedown or dissolve. Past 12 slots the default auto render strategy chunks the render, and single is refused above 12, so a 30-slot episode should leave render.strategy on auto. Sources must be Sume-hosted, so import generated files first; the same flow is in assemble long-form video with Timeline.

What if a shot runs long?

Cut it with video_trim: output length is 0.2 to 900 seconds, and the source can be up to 1800 seconds. start is required and exactly one of end or duration must be given. That is enough to cut one long take into several episode beats.

What about keeping characters consistent?

Consistency is a prompting and reference problem, not a Timeline one. See consistent characters for microdrama episodes for how to reuse references across shots.