TTS API pricing 2026: what 1M, 10M and 100M characters a month really cost

Most TTS pricing comparisons quote a rate per million characters and stop there. That rate is not what you pay: subscription fees, included usage and plan tiers change the monthly bill, and five different billing units are in use across the market. This post does the arithmetic for three real volumes, 1M, 10M and 100M characters in a month, picks the cheapest plan each provider sells at each volume, and puts each model’s blind-listening score beside its price.

The short version: among providers that publish a per-character price, SpeechifyAI is the cheapest at all three volumes, at $10, $91 and $631 a month on Simba 3.2. Simba 3.5, listed on 21 October 2026 after this round, is priced at its own rate, twice Simba 3.2’s (pricing). The next cheapest costs 1.5 times as much at 1M, 1.65 times at 10M and 1.9 times at 100M. And the cheap alternatives score far lower in blind listening, while Simba 3.2 was the lowest-priced model in the board’s top ten on 23 September and again on 13 October 2026. Evidence: the pricing snapshot of 28 September 2026, the leaderboard snapshot of 23 September 2026 and the 13 October reading.

Monthly cost at 1M, 10M and 100M characters

Sorted by the cost at 10M characters. Elo is the blind-listening score from the Artificial Analysis Speech Arena, read 23 September 2026 and kept as our reviewed snapshot of the full board.

Price against quality

The cheapest alternatives are also the lowest-rated. Azure Neural (1,031), OpenAI’s tts-1 (1,086) and Amazon Polly Neural (890) cost 1.6 to 1.8 times what SpeechifyAI does at 10M characters and score 150 to 350 points lower in blind listening. Evidence: the pricing snapshot of 28 September 2026 and the leaderboard snapshot of 23 September 2026.

Every model in the table rated above Simba 3.2 costs 1.9 to 4.1 times as much at 10M characters a month, on the same pricing and leaderboard snapshots. For how the top of the board compares on quality and price, see our TTS API comparison.

Token-priced models: Gemini TTS and gpt-4o-mini-tts

Google and OpenAI bill their newest TTS models per token, which cannot be turned into a per-character price from their own pages:

• Google Gemini 3.8 Flash TTS costs $0.50 per million text-input tokens and $9 per million audio-output tokens through 31 December 2026, then $1 and $18. Gemini 3.8 Flash-Lite TTS costs $0.50 and $6, then $1 and $12. Audio is metered at 25 tokens per second.
• OpenAI gpt-4o-mini-tts costs $0.60 per million text-input tokens and $12 per million audio-output tokens.

Neither vendor publishes how many characters a second of audio carries, so any per-character figure is an estimate. Artificial Analysis publishes one: $16.50 per million characters for Gemini 3.8 Flash TTS and $11.00 for Flash-Lite, in line with Google’s 2027 rates (23 Sep snapshot).

Free tiers

SpeechifyAI’s plans at each volume

Each SpeechifyAI plan is a monthly fee that buys a usage credit, then a per-character rate beyond it. The cheapest plan changes with volume:

Starter is the cheapest plan up to about 10.8M characters a month, Pro ($99, 13.5M characters included, then $8 per 1M) from there to about 63.5M, and Scale above that. Plans also differ in how many requests can run at once, and TTS API concurrency and rate limits by plan compares them across providers. The free plan covers 500,000 characters a month and pauses at the cap. Full details are on the pricing page, and how TTS pricing works explains per-character billing.

How each figure was built

For every provider and volume we took the vendor’s published plans, added the subscription fee to the usage beyond what the plan includes at that plan’s rate, and kept the cheapest result.

At 100M the plans change: Azure’s 80M commitment tier ($960, then $12 per 1M) comes to $1,200, Inworld’s Developer plan ($300, then $15 per 1M) to $1,500, Deepgram’s Growth rate ($27 per 1M, prepaid yearly) to $2,700, Cartesia’s Scale plan ($299, then $38 per 1M) to $3,795, and SpeechifyAI’s Scale plan to $631. Alibaba’s own model page lists Qwen-Audio-3.0-TTS-Plus at $20 per million characters; Artificial Analysis’s estimate is $27.60. Enterprise contracts, taxes and regional price differences are out of scope.

Start with the cheapest one that sounds right

Price only matters once the voice is good enough for your product, so listen before you commit. For a voice agent, weigh latency too: see how to choose a low-latency TTS API and WebSocket TTS streaming for voice agents. You can start free with 500,000 characters a month, no card, and run your own text through Simba 3.2 before you pay anything.