50,000-character voiceover: three Sume TTS requests, $2.375

A 50,000-character script costs $2.375 on Sume TTS (50 x $0.0475) and has to be sent as at least three requests, because one request is capped at 20,000 characters, which is $0.95. The split 20,000 + 20,000 + 10,000 costs $0.95 + $0.95 + $0.475.

The cap and the split

The catalog lists text-to-speech at $0.0475 per 1,000 characters and a maximum request of $0.95, equal to 20,000 characters. A longer script therefore needs several requests. Because the price is per character, splitting changes the number of jobs but not the total, so cut at paragraph or sentence boundaries rather than at the character limit.

Practical details

Cut on paragraph breaks so voice and pacing stay natural across files, and keep each request comfortably under the cap in case your text-length counting differs from the service's. A script that is 19,000 characters of Latin text can still be close to the limit if it holds many non-ASCII characters, so measure with the same character count you use for billing.

Each request is its own job. Use a distinct Idempotency-Key per chunk and the same key only for an exact retry; the jobs docs describe polling for results. When you join the audio, remember that joining is separate: the timeline audio concat job is $0.01 per job, so three files joined once adds $0.01.

• Cost per hour of speech depends on how many characters you speak per minute; measure your own script.
• Total with the join: $2.375 + $0.01 = $2.385.
• Retrying a failed chunk costs only that chunk.

Checking the estimate

Before you submit a long script, compute the price yourself: characters divided by 1,000 times $0.0475. A script of 12,345 characters is 12.345 x $0.0475 = $0.5864, which is one request. A script of 20,001 characters is 20.001 x $0.0475 = $0.9500 plus a bit, and would not fit one request, so you would send two.

The catalog quotes a default estimate of 1,000 characters at 5 cents, which is a rounded figure for planning. Use the per-character rate for real budgets. For a run of many scripts, add the per-script totals and then add $0.01 for each audio join job you need.

Sources

Prices are from the public catalog. Job polling is in the jobs and results docs.