Museum audio guide: 80 stops in three languages, TTS cost

An audio guide with 80 stops of 700 characters each, recorded in three languages, is 168,000 characters. At list that is $2.52 on MAI-Voice-2.1-Flash, $3.70 on MAI-Voice-2.1 and $7.98 on Sume TTS. The whole guide costs less than a ticket to the museum, so voice price is not the budget risk.

The bill

The risk is the language setup: three languages mean three scripts, three voices and 240 files to track.

One voice per language

Sume TTS takes a language field per request and expects the voice to fit it. Plan one voice per language and keep it for the whole guide, so stop 1 and stop 80 sound the same. If the voice and the language disagree, Sume returns a tts_voice_language_warning, so read the result's warnings after the first job instead of finding the problem in the gallery.

A 700-character stop is roughly 45 seconds of speech by a rough estimate, so 20 stops add up to about 900 seconds, under the 1,200-second limit. Twenty stops is 14,000 characters, under the 20,000-character cap too. That gives four jobs per language, 12 in all, each returning one hosted file per stop when you ask for sentence slices in a wav container.

Languages and file naming

Microsoft lists 23 languages and 26 locales for MAI-Voice-2.1, so a wide guide is a real use case for it. If one of your three languages is outside the Sume voice list, that decides the vendor before price does. Check the language list on each side first.

Number the stops in the same order in every language. A lookup table keyed by stop number, language, job id and segment index gives you the file for any visitor's phone, and one extra column for the script version tells you what to re-record after an edit.

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.

Pilot with a native speaker

Run stop 1 in all three languages as a pilot. Play each file to a native speaker, because names of artists and places are where synthetic voices slip. Fix the script text or the spelling, then run the whole batch.

Budget a second pass. A reviewed guide usually changes a few stops, and at 14,000 characters per job a full re-run of one language costs about $0.67 on Sume at list. That is cheap enough to re-run any batch that a curator edits, so there is no need to patch single stops by hand.

Keep a short pronunciation list per language: artist names, place names and any title in a second language. Apply it to the script text before you submit, so the same fix carries into every later re-run. Review the list with the curator, because a name that sounds right to a developer may not to a visitor from that region.

Do the same count for a later fourth language. Only that language's 56,000 characters are new, which is $2.66 on Sume at list, and the first three languages stay untouched.