Lip-sync audio under 5 seconds: join two TTS lines with audio concat

If your TTS line is under 5 seconds, it is too short for the H3 Max lip-sync route on Sume, which takes Sume-hosted audio of 5 to 14.8 seconds. Join it to the next line with POST /v1/timeline-1.0/audio using operation: "concat". The join is sample-exact with no re-synthesis and costs $0.01 flat per job.

The window is in the lip-sync route description of the Sume API reference, and the concat rules are in the Timeline audio docs (both read 2026-10-06).

How do I join the lines?

Generate each line with TTS as WAV (output_format wav, pcm_s16le) so the join stays sample-exact; the docs say to keep WAV if the file drives lip-sync. Then send parts[] with up to 20 files, each { url } with an optional source_in and duration. All the URLs have to be audio on this workspace's media.sume.com and have the same channel layout, and an Idempotency-Key is required.

• The result is kind: timeline_audio with one audio_url, duration_seconds and segments[].
• Check that duration_seconds is at least 5 and at most 14.8.
• Pass the new audio_url and a duration_seconds in 5 to 14.8 to the lip-sync route.

What does it cost?

The join is $0.01. H3 Max is billed per second of audio. Check the live rate for your resolution in GET /v1/catalog before you queue a batch. A line that is a little too short is cheaper to pad with the next sentence than to send to a model that refuses it.

What if it is over 14.8 seconds?

The API refuses a duration_seconds outside 5 to 14.8 and never clamps it. Split the audio at a sentence end with the same endpoint and operation: "split", and send each part to the lip-sync route separately.