Weekly podcast, 3 clips an episode: 156 captioned clips cost $34.32

A year of a weekly podcast at three vertical clips per episode is 156 clips. On Sume, trim ($0.02) plus captions ($0.20) is $0.22 a clip, $34.32 for the year.

5 min readSume
All posts

Three captioned clips from each weekly episode, for 52 episodes, is 156 clips. On Sume each clip is one video trim at $0.02 plus one standalone caption job at $0.20, which is $0.22 a clip. The year costs 156 x $0.22 = $34.32. Add an optional audio detach at $0.01 per clip if you transcribe elsewhere, and the figure becomes 156 x $0.23 = $35.88.

The per-clip chain

Each clip is two public-rate jobs on the same file. First, video trim cuts a [start, end) range from the episode into a new MP4 with precision: "exact". Then video captions burns captions onto that clip. The caption job reserves and captures $0.20 for videos of up to 60 seconds under the current fixed estimate, so keep clips at or under a minute.

The trim output is limited to 900 seconds and the source to 1800 seconds. A full episode longer than 30 minutes will not fit as a single source, so cut it into halves first, or trim from a shorter export.

Podcast clip jobs and public rates, as of 2026-10-08
JobRateLimits from the docs
Video trim$0.02 per jobSource up to 1800 s, output 0.2 to 900 s
Video captions$0.20 per jobFixed estimate covers videos up to 60 s
Audio detach (optional)$0.01 per jobSource up to 1800 s, output up to 900 s

Yearly arithmetic

The table uses 3 clips a week for 52 weeks. Change the first column to your own cadence.

Yearly cost by clips per episode, as of 2026-10-08
Clips per episodeClips per yearCost per clipYear total
152$0.22$11.44
3156$0.22$34.32
5260$0.22$57.20

Where the speech needs care

Captions made from speech need audible speech. If a clip is only music, the caption job fails with caption_no_speech, and the suggested next action is use_overlay_captions. For a cold open that is music only, send authored cues (text with start and end) instead of asking for speech recognition.

Pass script_text when you have a transcript you trust. Sume keeps the speech-to-text word times and aligns your words to them, which fixes names and product terms that recognition gets wrong. If alignment fails, the job returns script_alignment_mismatch or script_alignment_failed, and the advice is to simplify the script or omit it.

Idempotency per clip

Both jobs need an Idempotency-Key. Build it from the episode and the clip, for example ep212-clip2-trim, so a retried request does not queue and bill a second cut. The key records the job, so a replay returns the same job and not a new one.

Why the cadence matters for the budget

The rate does not change with volume, so the cost is linear in clips. The real lever is clip length: a caption job is priced for videos of up to 60 seconds, and most podcast clips are shorter. Sixty seconds is also a good upper bound for a vertical clip, so staying under it keeps both the caption price and the viewer on your side.

If you want to know the budget before you start, multiply clips by $0.22, and then add whatever you spend on picking moments. A spreadsheet row per episode with the three start times and durations is enough to drive the whole year from a script.

What this does not include

The $34.32 covers trim and captions only. It does not include picking the moments, which you do from a transcript or by listening, nor a B-roll insert or a hosted upload step. Import the episode once with POST /v1/media-imports, because trim reads only media.sume.com files of your workspace.

Sources

Related posts

More in Use cases

All Use cases posts

Written by Sume