Live-sale replay to 10 highlight clips: import, trim, caption

To turn one live-sale replay into ten highlight clips on Sume, import the replay once for $0.15, trim ten ranges at $0.02 each and caption each clip at $0.20, which comes to $2.35 in total. The one condition is that the replay must be a public TikTok or Instagram video URL.

That condition is real and worth stating first. Sume's import endpoint only accepts public TikTok or Instagram video URLs. It rejects YouTube and other hosts, and it does not take a file from your computer or a recording from a live-shopping app. If your replay lives somewhere else, this workflow does not apply as written.

Step 1: import once

POST /v1/media-imports mirrors a public TikTok or Instagram video into your workspace and costs $0.15 per import. After that the clip has a media.sume.com URL, which is what the media tools require. You must send an Idempotency-Key. See the media inputs docs.

Import only a replay you have the right to reuse, which in this case is your own posted replay.

Step 2: trim ten ranges

Video trim cuts a [start, end) range out of one Sume-hosted clip into a new MP4. It costs $0.02 per job and the output can be up to 900 seconds. Send start plus either end or duration. Pick moments by listening to the replay: a product reveal, a price drop announcement, a question answered. Ten trims are 10 x $0.02 = $0.20.

The source does not change, so one import feeds all ten trims.

Step 3: caption each clip

Each trimmed clip has a Sume-hosted URL, which works as the video_url for video captions at $0.20 per job for clips up to 60 seconds. Because live sales are spoken, the speech-to-text path will pick up the words. Ten clips cost $2.00. Keep the highlights under 60 seconds.

What this does not do

Sume does not connect to live-shopping apps, pull a stream while it is live, or detect the best moment for you. You choose the ranges. If you need help finding moments, read the replay's comments and your sales log for timestamps. The older video analysis endpoint is retired in the docs, so do not plan around it.

Captions will transcribe what was said, including prices. Check that every price in a clip is still valid before you post it, because a highlight clip stays online after the offer ends.

Picking the ten moments

Highlights that work are self-contained. A viewer who lands on the clip cold should understand what is being sold, see the item, and hear the price or the reason to care within the first seconds. Start the range a beat before the host picks the item up and end right after the key sentence. Resist the urge to include the whole segment, since a 20 to 40 second clip is easier to finish.

Keep a simple log with the replay time, the item, the price said aloud and the caption job reference. When an item sells out, you can find and retire its clip quickly. Exact trims are billed the same as keyframe-aligned ones, so choose by accuracy; the keyframe versus exact post explains the difference.

Run the import, wait for the job to finish, then queue the ten trims with unique idempotency keys. Poll GET /v1/jobs/{id}/status for each and fetch the results with /result. Only caption the clips you plan to post.