Daily 45-second news brief: a year of TTS on MAI Flash vs Sume

A daily brief that runs 45 seconds is about 700 characters, which is my estimate for normal speech, so a year of 365 briefs is 255,500 characters. At list that costs $3.83 on MAI-Voice-2.1-Flash, $5.62 on MAI-Voice-2.1 and $12.14 on Sume TTS. Either way, a year of daily audio is a rounding error in a publishing budget.

The bill

Flash makes up to 45 seconds of audio per call, so a 45-second brief sits right on its ceiling. That edge is more important than the price.

Writing to the ceiling

A brief written to exactly 45 seconds has no room. If a day's speech comes out at 46 seconds on Flash, the call hits the ceiling. Write to about 40 seconds, or about 620 characters at the same pace, and measure a few real briefs to learn your own ratio of characters to seconds.

On Sume a 700-character brief is one tiny job, far under the 20,000-character cap and the 1,200-second limit, so the length limit is not the issue. The result gives a duration in seconds, so a scheduler can reject a brief that came out too long before it publishes.

The scheduled job

The daily pattern is a scheduled job: write the text, submit a TTS job with an idempotency key built from the date, poll until terminal, then publish the audio_url. A repeat run on the same date with the same key does not bill a second time, so a flaky cron job cannot double the bill.

Put the brief on video too. A 45-second brief fits a vertical short, and you can send the TTS audio as the spine of a Timeline 1.0 render with captions. Captions are billed separately, so budget them next to the voice.

What the table leaves out

Microsoft's figures are the launch prices from 2026-10-01 (Microsoft AI, read 2026-10-05): $22 per 1M characters for MAI-Voice-2.1 and $15 per 1M characters for MAI-Voice-2.1-Flash, which makes up to 45 seconds of audio per call. Sume's figure is its list rate per transcript character, and on Sume spaces and punctuation count toward usage, so count characters of the final text.

The cheaper per-character price does not make the Microsoft route a drop-in swap. A Sume TTS request is an async job whose result carries a hosted audio_url, duration, optional word timings and sentence slices, and that hosted file is what Timeline audio and renders accept. Count the extra work of storing and joining files outside Sume before you call the gap a saving.

Check language coverage first. Microsoft lists 23 languages and 26 locales for MAI-Voice-2.1, and Sume takes a language field that you must set for every non-English transcript. A cheaper model that does not speak your target language is not an option, so settle that before you compare totals.

Fix the voice, log the duration

Pick one voice and one speed for the whole year and write them into the job template. Listeners notice a changed voice on a daily show far more than a small price difference.

Check the first week by ear and by clock. Log the duration of each brief, and if the average drifts toward 45 seconds, cut the script, not the speed.

A year is also long enough for a script style to creep. Read the character count of every brief and plot it monthly. A brief that grows from 650 to 800 characters raises the yearly bill by a few cents but pushes a Flash call past its ceiling, so the trend line is a better early warning than the invoice.

If you also publish the brief as text, reuse that same text for the audio so the two never disagree. One source, one job per day, and the audio file is a pure function of the text.