Text to Speech Pricing Compared: 2026 Price Guide
Text to speech pricing in October 2026: per-million-character API rates, subscription cost per minute, free tiers and billing gotchas across a dozen providers.
Text to speech pricing is hard to compare because providers bill in characters, tokens, credits and minutes. As of October 17, 2026, list API prices run from $4 to $100 per million characters, and most neural voices sit between $15 and $30. Converted at about 900 characters per minute, that is roughly $0.004 to $0.09 per minute.
How to compare TTS prices
Each provider uses its own unit. APIs usually charge per character. Some models, such as Gemini's speech models and OpenAI's gpt-4o-mini-tts, charge per token. Subscriptions use credits, and OndaLabs, ElevenLabs and Cartesia all count one credit per character on their standard models. Many subscriptions quote minutes instead.
We convert every figure to a cost per minute using 900 characters of speech per minute, which is 150 words at 6 characters per word. Real voices vary with language and pace, and Microsoft lists speaking speeds from 139 to 293 words per minute across its voices. Fish Audio bills by UTF-8 bytes, so accented or non-Latin text counts for more per character. Treat every derived figure as an estimate.
Pay-as-you-go API prices
The table lists API list prices as of October 17, 2026, with the derived cost per minute at 900 characters per minute. Where a vendor publishes its own per-minute figure, we use that instead.
| Provider and tier | Price | Derived cost per minute | |---|---|---| | Amazon Polly Standard | $4 per 1M characters | $0.0036 | | Amazon Polly Neural | $16 per 1M characters | $0.0144 | | Amazon Polly Generative | $30 per 1M characters | $0.027 | | Amazon Polly Long-Form | $100 per 1M characters | $0.09 | | Azure AI Speech Neural | $15 per 1M characters | $0.0135 | | Azure Neural HD | $22 per 1M characters | $0.0198 | | Google Chirp 3: HD | $30 per 1M characters, first 1M free | $0.027 | | OpenAI tts-1 | $15 per 1M characters | $0.0135 | | OpenAI tts-1-hd | $30 per 1M characters | $0.027 | | Deepgram Aura-2 | $0.030 per 1,000 characters | $0.027 | | ElevenLabs v4 | $0.08 per 1,000 characters | $0.072 | | ElevenLabs v4 Turbo | $0.04 per 1,000 characters | $0.036 | | Fish Audio s2.1-pro | $15 per 1M UTF-8 bytes | About $0.021 (vendor's 12 hours per 1M bytes) | | Murf Falcon API | $0.01 per minute | $0.01 (published) | | Gemini 3.8 Flash TTS | $9 per 1M audio output tokens to December 31, 2026; $18 from January 1, 2027 | $0.0135 now; $0.027 from 2027 (1,500 tokens per minute) |
As of October 17, 2026, Gemini's pricing page (last updated October 16, 2026) also lists $0.50 per 1M text input tokens for Flash and says the audio rate is 25 tokens per second. OpenAI's gpt-4o-mini-tts costs $0.60 per 1M text input tokens and $12 per 1M audio output tokens, and its page gives no per-minute estimate.
Subscription plans compared by cost per minute
Subscriptions bundle credits or minutes into a monthly fee. The table uses the vendor's stated minutes and excludes promotions. Prices as of October 17, 2026.
| Plan | Monthly price | Vendor-stated allowance | Derived cost per minute | |---|---|---|---| | ElevenLabs Starter | $6 | 30,000 credits, about 30 minutes | About $0.20 | | ElevenLabs Creator | $22 | 121,000 credits, about 121 minutes | About $0.18 | | ElevenLabs Pro | $99 | 600,000 credits, about 600 minutes | About $0.165 | | Cartesia Pro | $5 | 100,000 credits, about 133 minutes | About $0.038 | | Cartesia Startup | $49 | 1.25M credits, about 1,667 minutes | About $0.029 | | WellSaid Starter | $19 monthly, or $10 a month billed yearly | 20 minutes a month, or 240 a year | About $0.95, or $0.50 | | WellSaid Pro | $49 monthly | 180 minutes a month | About $0.27 |
Fish Audio and Murf plans are not converted here, because their units or allowances are not comparable from the pages we could read. ElevenLabs shows a first-month promotion on its Starter plan, which we excluded.
Free tiers and commercial use
A free tier is useful for testing. It is rarely the right place to publish commercially.
| Provider | Free allowance | Commercial use on the free tier | |---|---|---| | ElevenLabs | 10,000 credits, about 10 minutes | No | | Cartesia | 20,000 credits, about 27 minutes | No | | WellSaid Labs | Trial: 10 minutes of generation, 3 projects | No; paid plans have full commercial rights | | OndaLabs | 2,000 credits, about 2 minutes | Commercial use starts at Starter | | Google Chirp 3: HD | First 1M characters free | Not covered here | | Gemini TTS models | Free tier for both models | Check the terms before publishing |
Fish Audio's pricing page conflicts on its free plan, so we do not state whether that plan permits commercial use. Azure's free allowance could not be verified, so we leave it out.
Billing gotchas that change the bill
- SSML and markup. Azure bills SSML markup except the
<speak>and<voice>tags, while Amazon Polly does not bill SSML tags. - Chinese characters. Azure counts each Chinese character as two characters.
- Billing blocks. Twilio bills Say in 100-character blocks, with a 100-character minimum per charge.
- Tokens instead of characters. Gemini and OpenAI's audio models bill tokens, so a minute of speech is a token count rather than a character count.
- Price steps. Gemini's audio output price rises on January 1, 2027, and Fish Audio's older s1 model retires on December 31, 2026.
- Currency and region. Azure prices differ by region and currency, so the East US list prices used here may not match your bill.
How OndaLabs plans compare
OndaLabs sells credits in monthly plans and top-ups. Credit systems differ across vendors, so compare per minute rather than per credit. Check the pricing page for current plan terms. Prices as of October 17, 2026, with minutes derived from 900 credits per minute on Onda Lite.
| Plan | Monthly price | Credits per month | Derived minutes on Onda Lite | What is included | |---|---|---|---|---| | Free | $0 | 2,000 | About 2 | Onda Lite, 5 saved clips | | Starter | $9 | 170,000 | About 189 | Onda Flash, image generation, commercial use | | Creator | $22 | 400,000 | About 444 | 3 custom voices, music generation | | Pro | $49 | 900,000 | About 1,000 | 10 custom voices | | Business | $199 | 3,750,000 | About 4,167 | 30 custom voices, priority support |
Yearly plans cost 10 times the monthly price. Top-ups are $5 for 100,000 credits and $24 for 500,000. On Starter, that works out to about $0.048 per minute on Onda Lite and about $0.071 on Onda Flash (derived). To try the models, open the text to speech tool. OndaLabs does not sell a pay-as-you-go API, so the per-million figures are equivalents for comparison, not a published rate.
How to choose by workload
- Occasional voiceovers. Start with a free tier to audition, then pick a monthly plan that covers your real volume.
- Regular publishing. A subscription with a monthly credit allowance is easier to budget than per-call billing.
- Telephony or app builds. Compare API list prices per million characters, and remember that the lowest list price is not the same as the best value once quality, languages and features are counted. Among the list prices in this guide, Amazon Polly Standard is the lowest at $4 per 1M characters, but it is the basic tier.
- Custom voices. Check which plans include voice cloning or design, because the allowance differs. On OndaLabs, custom voices come from the Creator plan and up, and voices are created in the app.
Frequently asked questions
How much does text to speech cost?
It depends on the unit. API list prices range from $4 to $100 per million characters as of October 17, 2026, and most neural voices sit between $15 and $30. Subscriptions usually come to about $0.03 to $0.20 per minute on the plans we checked, with exceptions for small plans.
Which text to speech API is cheapest per million characters?
Among the list prices in this guide, Amazon Polly Standard is the lowest at $4 per 1M characters as of October 17, 2026. It is the basic tier, so compare quality and languages before deciding. Lowest list price is not the same as best value.
How many characters is one minute of speech?
This guide assumes about 900 characters per minute, which is 150 words at 6 characters per word. Real voices vary with language and pace. Measure a draft render to get your own figure.
Is there a free text to speech with commercial use?
Few free plans in this guide include commercial rights. ElevenLabs, Cartesia and WellSaid's trial do not, and OndaLabs' Free plan is for auditioning, with commercial use from Starter. Check each plan's terms before you publish.
Why are credits, characters and minutes hard to compare?
Each vendor counts a different unit, and some count tokens or bytes. Converting everything to cost per minute, with a stated assumption about speech rate, makes the figures comparable. Even then, the result is an estimate, so verify against your own usage.