How is text-to-speech priced?
Short answer: Mostly per character of text you send, but some providers bill tokens, credits or monthly plans, so compare in each provider's native unit.
The five units
| Unit | Example (official page, read 2026-10-08) | How to estimate |
|---|---|---|
| Characters | Amazon Polly Neural $16 per 1M; OpenAI tts-1 $15 per 1M; Speakvora $0.70 per 1M | Text length × price per million |
| Tokens | OpenAI gpt-4o-mini-tts: $0.60 per 1M text tokens and $12 per 1M audio tokens | Depends on the audio generated; measure a sample |
| Credits | Cartesia: 1 credit per character; $5 plan includes 100K credits | Credits ÷ plan price |
| Minutes / hours | ElevenLabs agents $0.08 per minute | Audio minutes × price |
| Subscriptions | ElevenLabs plans from $6 a month | Plan price ÷ included allowance |
Compute a bill
Characters per month ÷ 1,000,000 × price per million. 100,000 notifications of 45 characters is 4.5 million characters: $3.15 at $0.70, $67.50 at $15 and $72 at $16.
What changes the bill
- Repeats: some APIs bill every request, even for the same sentence. Storing audio yourself is the cheapest repeat (check each provider's terms).
- Free allowances: a monthly allowance and a one-time credit are not the same thing.
- Promotions: compare list prices, not temporary discounts.
Real numbers: text-to-speech pricing compared · how Speakvora counts usage · free tiers compared