Sume TTS price steps: 210, 421 and 631 characters per job

A Sume text-to-speech job costs 1 cent up to 210 characters, 2 cents up to 421, 3 cents up to 631, 4 cents up to 842 and 5 cents up to 1,052. The price is the Sonic list rate (38 micro-dollars per character) times 1.25, rounded up to a whole cent for each job, so the cost rises in steps rather than smoothly.

Where the steps come from

One cent is 10,000 micro-dollars. At 47.5 micro-dollars per character, one cent covers 10,000 / 47.5 = 210.5 characters. A job of 210 characters is 9,975 micro-dollars, which rounds up to 1 cent. A job of 211 characters is 10,022.5 micro-dollars, which rounds up to 2 cents. The same logic repeats every 210.5 characters.

The same rate applies to sonic-3.6, sonic-3.5, sonic-3 and the sonic-latest alias in the catalog, so the table does not change with the model id.

100 short lines: the boundary matters

Take a batch of 100 lines. If every line is 210 characters, the batch is 100 x 1 cent = $1.00. If every line is 211 characters, the batch is 100 x 2 cents = $2.00. One extra character per line doubles the bill. Trimming a 240-character line to 210 saves a cent per take, and across a few thousand lines that is real money.

The reverse also holds. Because the cent is charged per job, merging two 100-character lines into one 200-character job costs 1 cent instead of 2. Merging only helps if you can cut the audio apart again; the Timeline audio split operation slices one file into up to 20 ranges, and each range can carry its own start and end.

A practical rule for scripts

Most lines in an ad or a lesson are under 210 characters. If yours are 215, you are paying double for five extra characters. Count characters before you send, and tighten the copy to the step below. For longer takes, the step is 210.5 characters wide, so a 400-character paragraph costs the same 2 cents as a 421-character one, and there is no gain in trimming it to 380.

Note that spaces and punctuation count as characters in the request, so count the raw text you send, not the words.

Checking the voice settings you paid for

A completed text-to-speech job records the engine (model_id), the voice, the language, the output format and the synthesis settings. Read them from the job to make the next line sound the same, rather than guessing which settings the first take used.

• Under 211 characters: 1 cent.
• Per job, not per minute of audio.
• The 20,000-character request cap applies on top of the steps; a full-size job is $0.95.