50 Shorts voiceovers a month at 600 characters: MAI Flash vs Sume

A channel that posts 50 Shorts a month with a 600-character script each speaks 30,000 characters a month. At list that is $0.45 on MAI-Voice-2.1-Flash, $0.66 on MAI-Voice-2.1 and $1.425 on Sume TTS, which rounds to $1.43. Even the dearest option costs less per month than a coffee, so price should not decide this one.

The bill

What should decide it is how the voice gets into the video. The numbers below only count characters.

What a month of Shorts looks like

YouTube Shorts can now run up to three minutes, so a 600-character script is a short one. On Sume each Short is one TTS job, 600 characters against the 20,000-character cap and far under the 1,200-second audio limit. You make 50 jobs, each with its own idempotency key, and each returns a hosted audio_url that a Timeline 1.0 render accepts as audio.url.

Ask for word timings in the same job. They cost nothing extra in the schema, and they let you caption the Short from the script instead of running speech-to-text. A standalone caption job is $0.20 for a video of up to 60 seconds, a separate line item that the voice price does not include.

Flash makes up to 45 seconds of audio per call, so a script that runs longer needs more than one call there. Measure one take before you assume the script fits either way.

Where the difference stops mattering

At 50 Shorts a month the voice is the cheapest step in the chain. The render, the caption and the music cost more than the narration, so a 90 cent difference between Flash and Sume is not a reason to run two pipelines.

The reasons that do matter are operational. One vendor for voice, timeline and captions means one job id per Short, one failure path, and no file transfer between services. If you already pass the voice through another tool for brand reasons, the extra step is the cost to compare, not the characters.

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.

Run a week first

Run the first week as a pilot: seven Shorts through the full chain, with the same script length. Record the TTS job duration and the final video length for each. If any Short comes out over the limit of the platform you post to, shorten the script or set generation_config.speed up to 1.5 before you scale to the whole month.

Keep one voice id for the channel so episodes sound alike, and store it with your settings. Sume returns the voice and speed back on the job result, so you can read them later instead of guessing.