Sume TTS one-cent floor ends at 211 characters (vs MAI-Voice)

Sume text to speech has a one-cent floor per job, and the per-character list price of $0.0475 per 1,000 characters only exceeds that floor at 211 characters (210.53 characters is the exact crossover, so 210 characters still bills 1 cent). Below that length a Sume job costs the same 1 cent whether the script is 3 characters or 210. MAI-Voice-2.1 lists $22 per million characters and MAI-Voice-2.1-Flash lists $15 per million, so neither has a visible floor on its model page, and both are cheaper than Sume at every length in this post.

The numbers at seven script lengths

The Sume column uses the larger of 1 cent and characters x $0.0475 / 1,000, from the catalog's stated minimum of 1 cent and its note that spaces and punctuation count. The Microsoft columns are plain characters x the page price, with no minimum applied because the page states none.

Where each price crosses one cent

Divide 1 cent by each per-character price: Sume 210.50 characters, MAI-Voice-2.1 454.50 characters, Flash 666.70 characters. That means a 300-character line costs $0.01425 at list math on Sume, $0.0066 on MAI-Voice-2.1 and $0.0045 on Flash. The catalog also says its billing unit is the cent and that submit-time reservations round up to cents, so for many tiny jobs read the estimate on a real request before you trust list math to the last fraction.

• Under 211 characters: Sume cost is flat at 1 cent, so batching lines into one job saves money only if the voice and settings are the same.
• Over 211 characters: Sume cost scales at $0.0475 per 1,000 characters up to the 20,000-character request maximum ($0.95).
• Microsoft's model page lists no job floor or request cap in the fields I read, so I do not state one.

What the floor means for a voiceover with many short lines

A 14-line voiceover where every line is 150 characters totals 2,100 characters. List math on that total is $0.09975, but 14 separate jobs bill 14 x $0.01 = $0.14 because each 150-character line sits under the floor. Sending the same 2,100 characters as one job costs about $0.0998. If you need separate files per line for a Timeline, the floor is the price of that separation; if you can cut the audio afterwards, one job plus Timeline audio split at $0.01 per job is the alternative, and the split is a sample-domain cut with no re-synthesis. For one 2,100-character job plus one split, the arithmetic is $0.0998 + $0.01 = $0.1098.

How to decide

Count the lines before you count the characters. If your workflow produces many sub-211-character jobs, price it as jobs x 1 cent on Sume and as characters x the per-million rate at Microsoft. If it produces few long jobs, the per-character gap dominates: Sume lists $47.50 per million characters against $22 and $15. The remaining reasons to pick Sume are in the surrounding workflow, such as the durable result, the job envelope and the hand-off to Timeline, not in the per-job price.