Sume TTS maximum cost per request: $0.95 at 20,000 characters

One Sume TTS request takes up to 20,000 characters at $0.0475 per 1,000, so the most one request can cost is $0.95. A cost ladder with the rules.

4 min readSume
All posts

What is the most a single Sume TTS request can cost? $0.95. The request transcript accepts 1 to 20,000 characters, and the catalog rate for TTS 1.0 is $0.0475 per 1,000 characters, so 20 thousand-character blocks come to $0.95.

Spaces and punctuation count toward usage, so count the whole string you send, not just the words. These limits are in the Sume OpenAPI spec.

A cost ladder

The rate is per character, so cost scales in a straight line. The catalog lists an estimate of 1,000 characters, a minimum of one cent and a maximum at 20,000 characters, expressed in whole cents in the catalog, which is why the displayed maximum is 95.

Sume TTS 1.0 cost by transcript length at $0.0475 per 1,000 characters (read 2026-10-03)
Characters sentCost at the list rate
500$0.02375
1,000$0.0475
5,000$0.2375
10,000$0.475
20,000 (request maximum)$0.95

Counting characters honestly

Count what you send. A script with stage directions, speaker labels or markup in it bills for those characters unless you strip them first. A good habit is to build the transcript in code, then log its length before the request.

For a non-English script, set language explicitly. The spec says an omitted language defaults to English at the provider, with a fallback that infers Korean or Japanese from a transcript of only Hangul or kana. A wrong language setting wastes the characters you paid for.

What else caps one request

There is a second ceiling besides characters: synthesized audio longer than 1,200 seconds fails with tts_duration_exceeded, and the spec states there is no credit capture on that failure. So a request can hit the duration cap before it hits 20,000 characters if the speech is slow.

Speed is settable from 0.6 to 1.5 in generation_config. A slow setting stretches the audio, so a long script at 0.6 is the likeliest way to reach the duration cap.

Splitting a long script

When a script is longer than one request, split on sentence or paragraph boundaries, send each block as its own request with its own Idempotency-Key, then join the audio. A sample-accurate join keeps seams clean; the timeline audio endpoint concatenates up to 20 parts into one file for a flat $0.01 per job.

Plan the number of blocks first: a 45,000 character script is three requests of 15,000 characters each, which is 3 times $0.7125, or $2.1375 at the list rate. The same rate applies whatever the block size, so splitting does not change the total.

Confirm the live rate

Rates come from the catalog, so read GET /v1/catalog (see the API reference) before you budget a large run. For waiting on results rather than holding a request, see Sume jobs and results. A related look at long scripts is 20 minutes of narration against the 20,000 character limit.

Sources

Related posts

More in Pricing

All Pricing posts

Written by Sume