Join TTS lines into one track: match output_format before concat

To join several Sume TTS lines into one voiceover track, send the same output_format on every line, then call timeline audio concat once with the audio urls as parts. Concat takes 1 to 20 parts and all of them must share a channel layout, or the job fails with audio_parts_channel_mismatch.

Match the lines

TTS 1.0 defaults to mp3 at 44100 Hz and 128 kbps. It also supports wav and raw. If every line uses the default, the parts match. Mix a mono and a stereo line and the concat fails. The cleanest rule is to set one output_format in your batch code and reuse it.

Longer scripts

Past 20 lines, concat in groups of 20 and concat the group outputs. Or skip the join: Timeline 1.0 takes audio.parts[] for a single render, so you need no separate audio file. Use concat when you need a durable track for another job, such as an avatar video's audio.

Poll the job

Timeline audio has no resource GET. Read GET /v1/jobs/:id/status, then /result. With mode: "sync" you wait at most 30 seconds, otherwise you get a 202.

A note on formats

The audio detach and timeline audio pages describe wav as the sample-exact choice, and mp3 as a smaller one at 128 kbps. When the track will feed another processing step, such as an avatar mux, a lossless line is the safer input. When it only goes into a render, mp3 is smaller to move around.

Whichever you choose, choose once. A mixed batch where some lines came back at a different rate or channel count is the usual reason audio_parts_channel_mismatch appears. The check runs on the worker, so you find out after the job is accepted.