The cheapest text-to-speech list price in my data is $4 per 1M characters, for Google WaveNet voices and Amazon Polly Standard. The most expensive is $100 per 1M characters, for MiniMax Speech 2.8 HD and ElevenLabs Multilingual v2 via fal.ai. Most providers bill per character, so I also show how to convert any price to a per-minute figure.
How to convert prices to one minute
Speech is usually sold by the character, but people budget by the minute. TaskCents uses about 150 words, or 900 characters, per minute of speech. The formula is:
- Cost per minute = (price per 1M characters) x 900 / 1,000,000
For example, a price of $15 per 1M characters works out to $0.0135 per minute. That matches Google's published per-minute rate for Gemini 3.8 Flash TTS.
Two caveats apply:
- A character count is not a time count. Speaking rate, pauses, voice, and language change how many characters fit into a minute. The 900-character figure is a planning assumption, not a measurement of any provider's output.
- Some providers publish per-minute or per-10-second prices. Where they do, I use their published figure.
List prices by provider
Official provider prices as of 2026-10-11, in USD, sorted from cheapest to most expensive.
| Service | Official price |
|---|---|
| Google WaveNet voices | $4 per 1M chars |
| Amazon Polly Standard | $4 per 1M chars |
| Gemini 3.8 Flash-Lite TTS (Google) | $0.009 per minute |
| Gemini 3.8 Flash TTS (Google) | $0.0135 per minute |
| OpenAI TTS-1 | $15 per 1M chars |
| Orpheus TTS (Together AI) | $15 per 1M chars |
| Azure Neural TTS | $15 per 1M chars |
| Google Neural2 voices | $16 per 1M chars |
| Amazon Polly Neural | $16 per 1M chars |
| Azure Neural HD TTS | $22 per 1M chars |
| Google Chirp 3 HD voices | $30 per 1M chars |
| OpenAI TTS-1 HD | $30 per 1M chars |
| Amazon Polly Generative | $30 per 1M chars |
| Deepgram Aura-2 | $30 per 1M chars |
| Cartesia Sonic-3.6 (Scale plan) | $37.375 per 1M chars |
| ElevenLabs Eleven v4 Turbo | $40 per 1M chars |
| ElevenLabs Flash / Turbo | $40 per 1M chars |
| Gemini 3.1 Flash TTS (via fal.ai) | $50 per 1M chars |
| MiniMax Speech 2.8 Turbo | $60 per 1M chars |
| Cartesia Sonic-3 (Together AI) | $65 per 1M chars |
| ElevenLabs Eleven v4 | $80 per 1M chars |
| ElevenLabs Multilingual v2 | $80 per 1M chars |
| ElevenLabs Eleven v4 (via fal.ai) | $80 per 1M chars |
| ElevenLabs Multilingual v2 (via fal.ai) | $100 per 1M chars |
| MiniMax Speech 2.8 HD | $100 per 1M chars |
The Cartesia Sonic-3.6 figure comes from its Scale plan: $299 per month for 8M credits, where one credit equals one character.
What stands out
- Budget tier: WaveNet and Polly Standard cost $4 per 1M characters. Google's Gemini 3.8 Flash-Lite TTS is $0.009 per minute.
- Mid tier: OpenAI TTS-1, Azure Neural, and Orpheus are $15 per 1M characters. Gemini 3.8 Flash TTS is $0.0135 per minute. OpenAI's HD tier is $30 per 1M characters, twice the standard tier.
- Premium tier: ElevenLabs Eleven v4 and Multilingual v2 list at $80 per 1M characters on ElevenLabs. MiniMax Speech 2.8 HD is $100 per 1M characters.
- Route matters: ElevenLabs Multilingual v2 is $80 per 1M characters on ElevenLabs and $100 via fal.ai. Eleven v4 is $80 in both places.
Per-character vs per-minute pricing
Per-character pricing is easy to compare across vendors, but the minute depends on your script, voice, and language. Per-minute pricing, which Google publishes for its Gemini 3.8 TTS models, is closer to what you pay for a finished file. It is still a list rate and does not guarantee how long your script will run.
Google lists the Gemini 3.8 Flash TTS and Flash-Lite TTS rates through December 31, 2026. Check whether the rate changes after that date before you build a long-term budget on it.
What the numbers leave out
The table covers list prices for synthesis only. Your real bill may also include:
- Free tiers and plan minimums. Cartesia's Scale plan, for example, is a monthly subscription, and Google Cloud voices have a free tier before the listed rate applies.
- Voice cloning or custom voice fees, which are not part of the list prices above.
- Retries, regenerated takes, and test runs.
- Storage and delivery of audio files.
If you also use a text model to write or rewrite the script, that cost is separate. Estimate it with the cost calculator, and see my methodology page for how I measure per-task costs for text models.
FAQ
Is 900 characters per minute accurate?
It is a planning figure, not a fixed rule. Speech rate, pauses, and language change the character count of a minute. Use it to compare providers, then test your own script to confirm.
Which option is cheapest?
Google WaveNet voices and Amazon Polly Standard, both at $4 per 1M characters. Among per-minute prices, Gemini 3.8 Flash-Lite TTS is $0.009 per minute.
Do premium voices cost much more than standard ones?
Yes. ElevenLabs Eleven v4 lists at $80 per 1M characters and MiniMax Speech 2.8 HD at $100, against $4 for WaveNet and Polly Standard. Whether the quality gap justifies the price depends on your use case.