TTS on Sume: 1,000 characters is 5 cents, the 20,000 cap is 95

The short answer

Text-to-speech on Sume costs 5 cents for a 1,000-character script, and 95 cents for the 20,000-character maximum a single request accepts. The rate is $0.000038 list per character times 1.25, which is $0.0000475 per character, and each job is rounded up to the next whole cent.

All five TTS Router engines (sonic-3.6, sonic-3.5, sonic-3, sonic-latest and sonic-preview) carry the same per-character price in the catalog, so the engine you pick does not change the bill.

Price by script length

The count is characters of the transcript, spaces and punctuation included. The micros column shows the unrounded billable amount in millionths of a dollar.

A budgeting rule you can use

The unrounded rate is 4.75 cents per 1,000 characters, or $47.50 per million. For a batch of short scripts, remember that rounding happens per job. Fifty separate 300-character scripts cost 50 x 2 = 100 cents ($1.00), although 15,000 characters in one request would cost 71.25 cents rounded up to 72. If you can combine scripts, one request is cheaper by about 28 cents here; if you need separate audio files per script, accept the rounding.

The request limit is 20,000 characters per job. A longer script has to be split, and each part is rounded up on its own: two 10,000-character jobs cost 48 cents each (47.5 rounded up), 96 cents in total, against 95 cents for one 20,000-character job.

Where this fits in a video budget

Narration is usually the smallest line in a finished video. A 450-character voice-over costs $0.03, against a Wan 3.0 clip at 720p for 10 s at $1.25. A request also needs a voice selector (an avatar with a ready voice, or a voice id), as the API reference describes. Timeline 1.0 then joins audio parts in the sample domain with no re-TTS.

The routes are listed in the API reference: GET /v1/tts-router/models returns the catalog with the pricing block, and POST /v1/tts-router/generate creates the job. For the full model families, see Models.