Revoice 20 of 120 training videos: pay for changed scripts only

When 20 of 120 training videos get a new script of 1,800 characters, re-voicing only those 20 is 36,000 characters: $0.54 on MAI-Voice-2.1-Flash, $0.79 on MAI-Voice-2.1 and $1.71 on Sume TTS. Redoing all 120 would be 216,000 characters, which is $3.24, $4.75 and $10.26.

The bill

Because speech is billed per character, the saving from redoing only the changed scripts is exactly the unchanged share, about 83 percent.

Know what changed

That only works if you know which scripts changed. Keep a hash of each script next to the job id of its last voice track, and on every content change re-hash. A video is re-voiced only when its hash differs, and everything else keeps its stored audio_url.

Each script is 1,800 characters, about two minutes of speech by a rough estimate, so ten scripts fit in a 20,000-character request and the 20 changed ones need two jobs. Ten scripts at two minutes is 1,200 seconds, right at the audio limit, so run nine per job to be safe or measure first.

Put it back and re-caption

A new voice track must go back into the video. If the video is a Sume Timeline 1.0 render, change audio.url to the new track and the render picks it up. Captions made from the old script no longer match, so re-run them for the changed videos too.

A word-level edit inside a long script can still be a big job. If one sentence changed in a 1,800-character script, you re-voice the whole script, because Sume voices the text you send, as a whole. Split long scripts into sections and voice each section as its own sentence slice, so an edit re-voices only its section.

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.

Keep a change log

A change log is worth more than the voice price. Record the script version, voice, speed and job id with every track. When a reviewer asks which training videos carry the old policy wording, the log answers in seconds.

Avoid mixing two voice vendors inside one course. A learner hears the switch, and a join of a Sume part with an outside file is not possible on a Sume timeline anyway.

Budget the first full rollout once, then treat re-voicing as a small recurring cost. At 1,800 characters a script, each re-voice is $0.0855 on Sume at list, so even a bad month with 30 edits is about $2.57. The review time of a person who checks the new narration costs far more than the speech does, so spend your effort on review.

Finally, keep the unchanged tracks' job ids. They are the proof that 100 of 120 videos were not touched, which is what a compliance reviewer wants to see when a policy changes.