A weekly 90-second recap video for a year: 39.5 cents each on Sume

A weekly 90-second recap costs 39.5 cents on Sume for the voice, a music bed and the render, which is $20.54 for 52 weeks, or $14.17 if you generate one music bed and reuse it. Add captions at 20 cents a week and the year is $30.94 with a new bed each time.

The weekly receipt

Assume a 1,350-character script, which is about 90 seconds at 15 characters a second. That rate is an assumption, so measure it from word timings on your first episode. The other numbers come from the catalog and docs.

Where the money goes

The render is half the bill. Timeline 1.0 bills $0.10 per ceil output minute, so a 90-second video is billed as two minutes, and trimming the recap to 60 seconds halves the render line to 10 cents. That one edit saves $5.20 a year, more than the voice costs.

Captions come with a limit worth reading twice: the 20-cent price is stated for videos up to 60 seconds, and a 90-second recap is longer. Confirm the live price for your length in the catalog before you put it in a budget.

Comparing the voice line

Microsoft's MAI-Voice-2.1 page lists Standard at $22 per million characters and Flash at $15 per million characters (read 2026-10-05). At 1,350 characters a week that is about 2.97 cents and 2.03 cents before any rounding or tax on that side, or $1.54 and $1.05 a year. Sume's voice line is $3.64 for the year. The difference is under $3, so for this video the voice engine is the least important line. Choose it on quality and on the voices you are licensed to use.

Three ways to cut the bill

• Reuse a bed. Generate three or four beds a year and rotate them, and the music line drops from $6.50 to about 50 cents.
• Shorten the recap to 60 seconds. The render line falls from 20 to 10 cents.
• Batch the planning. The compile preflight, POST /v1/timeline-1.0/plan, is unbilled, so check the output length there before any paid render.

What the table leaves out

It leaves out your time and the footage. If the weekly picture comes from stock or from screen capture, the video is already yours, and the receipt above is the whole spend. If you generate new video clips each week, that line will dwarf these three, so price it from the model's own page before you commit.

A note on the 15 characters a second assumption

The whole table hangs on 1,350 characters being 90 seconds. If your recaps read slower, the characters fall and the voice line drops with them, but the render line is billed in whole minutes, so it only changes when the video crosses a minute boundary. Measure the first episode with word timings and write the true character count at the top of your budget sheet. If episode one runs 1,100 characters, the voice line is 6 cents, not 7, and the year is $0.52 cheaper on that line alone.

Keep the model of the year in a spreadsheet and update the three inputs once a quarter. The prices on these pages are read dates, not promises, and the catalog is the source of truth for what a job costs on the day you run it.