silence_split_seconds: tune caption line breaks from STT segments

On Sume, sentence segmentation returns segments[] with no gaps, in the shape of caption lines, and silence_split_seconds sets how long a pause must be before it splits a line. The video inspect page gives the range as 0.2 to 3. A short value breaks lines at small pauses and gives more, shorter cards. A long value gives fewer, longer ones.

Where to set it

The setting is part of segmentation on a transcribing request. On video inspect you send transcribe: true, then segmentation.mode: "sentence". If you send language_code, segmentation or duration_seconds without transcribe: true, the API returns 400 video_inspect_transcribe_required.

{
  "video_url": "https://media.sume.com/artifacts/artf_demo/talk.mp4",
  "transcribe": true,
  "segmentation": { "mode": "sentence", "silence_split_seconds": 0.6 }
}

Choosing a value

The table says what the setting does in direction only. The docs do not give a recommended value, so test one clip and read the segments.

From segments to captions

Segments are time ranges, not sliced audio files. To burn them, map each segment to a cues entry with text, start and end, and send those to POST /v1/video-captions. Sume then burns your text at those times without running speech-to-text again. Send only one of script_text, words, cues and segments.

Transcribe adds the STT rate of $0.01 per audio minute to the inspect reservation. Without duration_seconds, Sume reserves 1 minute, and the maximum hint is 600 seconds.