Weekly Twitch highlight budget: 3-hour VOD to 6 clips

A weekly stream highlight workflow on Sume costs $3.28 in per-minute and flat lines: $1.80 to transcribe three hours of audio, $0.12 to trim six moments, $1.20 to caption them, $0.06 to detach their audio and $0.10 to render one reel. Over four weeks that is $13.12. That total leaves out the media import of each VOD piece and the compute that every video inspect also reserves, which the docs do not give as a flat rate, so check both in GET /v1/catalog before you budget.

The transcript is the biggest line, and it is the only part that scales with stream length. Trimming, captioning and rendering scale with how many highlights you ship.

How is the transcript priced?

Video inspect transcripts are billed at $0.01 per audio minute, and a transcript job takes a duration hint of up to 600 seconds. A 3-hour VOD is 180 minutes, so this plan uses 18 pieces of 10 minutes. Cutting the VOD into pieces and importing them to media.sume.com is your own step, and it is required: both video inspect and video trim accept a source of at most 1,800 seconds, so a 3-hour file cannot go in whole. Each highlight must also sit inside one piece. Per the docs, if you omit the duration hint the job reserves one minute, so send the real length.

Where the plan bends

Stream length. A 6-hour stream doubles the transcript line to $3.60 and moves the weekly total to $5.08. Everything else holds.

Clip count. Each extra highlight adds a trim, a caption and an audio detach: $0.23. Ten highlights instead of six adds $0.92 to the week.

Captions. The $0.20 caption price covers a video up to 60 seconds, so a highlight kept under a minute stays at one caption job.

A practical rule

Transcribe once per stream, pick highlights from the text, and only then pay for trims. Reversing that order buys captions for clips you drop. Keep a small reserve in the wallet for a retry of any single step; the largest single line in this plan is a $0.10 transcript job, so a reserve of a few dollars covers a full rerun of one piece several times over.

Monthly view and top-up

At the weekly total the month is $13.12, so a $15 top-up covers four weeks of these lines with a small cushion, and a $20 top-up leaves room for a double-length stream or two, before import and inspect compute. The dashboard can start a manual top-up when billing is configured for the workspace, so choose an amount you will actually repeat each month.

Because holds are reserved at submit and captured on success, the wallet balance can look lower than your total during a busy hour of jobs. The credits docs describe the hold and settle model; check there before you read a dip in the balance as spend.

• Transcript: $0.01 per audio minute, so $0.10 per 10-minute piece.
• Trims: $0.02 per job, so cutting more candidates than you ship is cheap.
• Detach: $0.01 per job gives each clip the audio file the reel render needs as its spine.
• Reel render: $0.10 per output minute, rounded up.

Cheaper variants

If you only need the text of the last hour, transcribe the last hour. If your highlights come from chat timestamps or clip markers rather than speech, skip the transcript entirely and pay only for trims, captions, audio detach and the render: $1.48 a week. The transcript is worth paying for when you want search over what was said.

A stream that runs 3 hours but has only 90 minutes of speech still bills by the audio you send. Trim dead air before you transcribe if your own tooling can do it.