A 25-minute episode into eight teaser clips: four cents of audio

A 25-minute episode becomes eight teaser audio clips for 4 cents on Sume: two audio-detach jobs of 750 seconds each, because detach outputs at most 900 seconds, then two timeline-audio splits of four ranges each. Each split job is a flat $0.01.

Why two detach jobs

The audio detach page allows a source of up to 1,800 seconds, and the output is limited to 900 seconds. A 25-minute episode is 1,500 seconds, so a whole-track extract fails and needs a range. A range longer than 900 seconds returns audio_detach_range_empty. Splitting the episode at the midpoint gives two ranges of 750 seconds that each fit.

The same page says that for many ranges from one track you should detach once and then split with timeline audio. That is the rule here, applied twice because of the cap.

The two request shapes

First the detach, with the range on the second half of the show. Idempotency-Key is required, and the video must already be on media.sume.com.

curl -X POST https://api.sume.com/v1/audio-detach \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: ep-part-2" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/episode.mp4",
    "range": { "start": 750, "end": 1500 },
    "mode": "sync"
  }'

Then the split, with ranges in seconds on the detached wav. Each range is start and an optional end, up to 20 per job, and each segment returned has its own audio_url.

curl -X POST https://api.sume.com/v1/timeline-1.0/audio \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: ep-split-2" \
  -d '{
    "operation": "split",
    "url": "https://media.sume.com/artifacts/artf_demo/part2.wav",
    "ranges": [{"start": 40, "end": 70}, {"start": 190, "end": 220},
                 {"start": 400, "end": 430}, {"start": 610, "end": 640}]
  }'

Choosing the ranges

The timestamps of the teasers are your editorial call. If you want them chosen from the content, transcribe the episode first: STT bills $0.01 per audio minute, with a duration hint of at most 10 minutes per request, so a 25-minute show is three or more calls and about 25 cents. The sentence segments from that transcript give you clean cut points that begin and end on a sentence.

Ranges in the second split are relative to its own file, which starts at zero. If you picked a teaser at 1,010 seconds in the episode, that is 260 seconds in the second wav. Keep a small table of episode time, file and local time, and the mistake never happens.

Limits and edges

• Wav is sample-exact on a split. A mp3 source adds priming padding at each edge.
• Ranges may overlap, which suits teasers that share a quote.
• A video without audio fails with detach_source_has_no_audio, so probe first if you did not record it yourself.
• Teasers are audio. Putting a picture to each is a separate Timeline 1.0 render at $0.10 per output minute.

What this does not do

The result is audio, not video. If each teaser also needs picture, you cut the video on the same timestamps in a Timeline 1.0 render, which is billed by output minute, so eight 30-second teasers are eight renders of one billed minute each, $0.80. The 4 cents buys the audio spine of the whole campaign.

Keep the source episode untouched on media.sume.com. Detach never modifies it, so a second pass with other ranges next week costs the same 4 cents and needs no re-import.

Checklist

Six lines, four cents, and the archive of the show is reusable for any later campaign without paying again.

• Import the episode with a media import so the URL is on media.sume.com.
• Probe for an audio track before paying for either detach.
• Detach two 750-second ranges with an Idempotency-Key each.
• Pick teaser ranges from a transcript or from your editor's markers.
• Convert episode time to local file time before you split.
• Store the eight audio_url values with their episode timestamps.