Text to speech API cost per minute of audio: $0.036-0.038 on Sume

The price is per character, not per minute

Sume TTS 1.0 and the TTS Router both list at $0.0475 per 1,000 characters, with a minimum of one character and a maximum of 20,000 characters per request. Nothing is billed per second of audio, so the cost of a minute depends on how many characters the voice speaks in that minute.

Cartesia's pricing page states that one minute of audio generation requires 750-800 credits and that one credit equals one character. Using that as the conversion gives a range for a normal pace.

Cost per minute

750 characters is $0.035625 and 800 characters is $0.038. Round that to 3.6 to 3.8 cents per minute of finished audio.

What moves the number

The character count is the invoice line, so anything that changes characters changes the price. Spaces and punctuation count, and the count matches the JavaScript string length, so a script full of markup, stage directions, or repeated headings costs more than the same words stripped clean.

• Strip stage directions and speaker labels before you submit.
• Do not send the same paragraph twice while testing; render one short sample first.
• A failed job that exceeds 1,200 seconds does not capture credits, but a successful one does.
• Changing speed changes duration, not characters, so it does not change the price.

Comparing to a vendor plan

Vendor plans price a bundle of characters, so the effective rate depends on how much of the bundle you use. Cartesia lists overage per million credits at $65 on Pro, $45 on Startup and $38 on Scale. Sume's $0.0475 per 1,000 characters equals $47.50 per million, a flat rate with no bundle to under-use. The overage comparison lays it out in a table.

A quick estimator you can paste into a spreadsheet

The whole model is one multiplication: characters times 0.0475 divided by 1,000. For minutes, multiply minutes by your characters per minute first. With the vendor's 750 to 800 range you get a low and a high estimate, and the true figure for your voice sits between them or near them.

Do not forget the other line items in a video pipeline. Speech is one job; captions are $0.20 for up to 60 seconds, audio detach and Timeline audio are $0.01 each, and music is $0.125 per generation. The speech is usually the cheapest part of a short ad.

Budget a batch

For a 50-episode podcast of 12 minutes each at 800 characters per minute, the arithmetic is 50 x 12 x 800 = 480,000 characters, or $22.80 at list price. Run one episode first to measure your real characters per minute, then scale. The API pricing page holds the current book.