Voiceover longer than your clips: Timeline pads, loops, 0.5 s rule

When a voiceover is longer than the clips you cut under it, a Sume Timeline 1.0 render still runs for the length you declare in audio.duration_seconds. Short sources are padded or looped and reported as soft warnings, not failures. But the video[] slots themselves can stop at most 0.5 seconds before the end of the voiceover, so you have to plan slot lengths that reach the end.

This follows the Timeline 1.0 docs (read 2026-10-06).

What decides the length of the output?

audio.duration_seconds is the output length. It is required and runs 1 to 1,800 seconds. Take it from the TTS result: with timestamps.words on, the last word's end gives the spoken length, and the result also reports duration_seconds. The render is priced at $0.10 per ceil output minute, and the reserve is ceil(audio.duration_seconds / 60) minutes.

What happens to a short clip?

The docs say soft warnings include padded or looped short sources, snapped transitions and ignored still motion. They are not failures. The render finishes, so check warnings[] in the result and look at the output before you publish.

How do I plan the slots?

Add up the slot durations first. Timeline has a plan step that checks URLs and compiles the program without creating a job or reserving credits; it returns duration_seconds, segment_count, billable_minutes and estimated_cost_usd_micros. Run it before the paid render. If you need to join several voiceover lines into one spine first, Timeline audio docs cover a gapless concat that returns segments[] offsets you can use to place the clips.