Free Text to Speech API: Real Limits Compared (2026)
Yes, you can get a free text to speech API, but "free" almost always means a monthly character allowance, not unlimited use. As of September 2026, Google Cloud, Microsoft Azure and the AtlasVoice API offer ongoing monthly free allowances, Amazon Polly ties most of its free tier to your first 12 months, ElevenLabs gives free credits for non-commercial use only, and OpenAI's pricing page lists no free tier for speech.
This guide compares the free tiers side by side, explains the fine print that decides whether a free plan works for your project, and gives an honest answer to the question "is there a truly unlimited free TTS API?"
What "free" means for a text to speech API
Before comparing numbers, it helps to know the five things that usually sit behind the word "free":
- A monthly character allowance. Almost every TTS API bills per character of text you send. The free tier is a number of characters each month, and spaces and punctuation usually count.
- A time limit. Some free tiers renew every month for as long as you have an account. Others only apply for your first months, after which every character is billed.
- A card or billing account. Several cloud providers require billing to be enabled even if you stay inside the free allowance, and they charge automatically once you go over it.
- Commercial-use rules. Some free plans only allow personal or non-commercial use, or require you to credit the provider when you publish the audio.
- Rate limits and missing features. Free tiers often allow fewer requests per minute, or leave out features such as batch synthesis or custom voices.
Keep these in mind as you read the table below. The biggest number is not always the best free plan for your use case.
Free text to speech API comparison (as of September 2026)
All figures come from each provider's official pricing or documentation page, checked in September 2026. Prices change, so confirm on the linked page before you build. AtlasVoice is our own product.
| API | Free allowance | How long | Good to know |
|---|---|---|---|
| Google Cloud Text-to-Speech | 4 million characters a month for Standard and WaveNet voices; 1 million for Neural2, Studio, Polyglot (Preview) and Chirp 3: HD voices | Every month | Billing must be enabled. No free usage for Instant custom voice or Gemini-TTS models. |
| Microsoft Azure AI Speech (Free F0) | 0.5 million neural characters a month | Every month | 20 transactions per 60 seconds, not adjustable. No batch synthesis or custom voice on F0. |
| Amazon Polly | 5 million Standard characters a month; 1 million Neural, 500 thousand Long-Form and 100 thousand Generative characters a month | Neural, Long-Form and Generative: first 12 months | New AWS customers since July 15, 2025 get up to $200 in Free Tier credits instead of a set allowance (see below). |
| ElevenLabs (Free plan) | 10k credits a month | Every month | No commercial license. Published audio must credit ElevenLabs. |
| OpenAI TTS | No free tier listed on the pricing page | Not applicable | Paid per use from the first character. |
| AtlasVoice API (our product) | 100,000 characters a month | Every month | No card needed. 30 requests a minute. Commercial use allowed. |
Sources (official pages, checked September 2026): Google Cloud Text-to-Speech pricing (opens in a new tab), Azure AI Speech pricing (opens in a new tab) and quotas (opens in a new tab), Amazon Polly pricing (opens in a new tab), ElevenLabs pricing (opens in a new tab) and publishing terms (opens in a new tab), OpenAI API pricing (opens in a new tab).
The free tiers in more detail
Google Cloud Text-to-Speech
Google's free tier renews every month: 4 million characters a month for Standard and WaveNet voices, and 1 million characters a month for Neural2, Studio, Polyglot (Preview) and the newer Chirp 3: HD voices. Instant custom voice and the Gemini-TTS models have no free usage.
The catch is billing. Google's pricing page says you must enable billing to use Text-to-Speech, and you are charged automatically once you pass the free characters. After the free tier, Chirp 3: HD costs US$30 per 1 million characters and Standard or WaveNet cost US$4 per 1 million. Setting up a key takes a few steps, which we walk through in how to get a Google Cloud TTS API key (opens in a new tab).
Microsoft Azure AI Speech
Azure's Free (F0) pricing tier includes 0.5 million characters a month of neural text to speech. According to Microsoft's quotas documentation, F0 resources are limited to 20 transactions per 60 seconds for real-time synthesis, and that limit cannot be raised. Batch synthesis and custom professional voice are not available on F0. It is a good fit for prototypes and low-traffic features. For production traffic you would move to the Standard (S0) tier, where the price depends on your region, so check the calculator on the Azure pricing page.
Amazon Polly
Polly's pricing page lists a free tier of 5 million characters a month for Standard voices, plus 1 million Neural, 500 thousand Long-Form and 100 thousand Generative characters a month for the first 12 months. After that, Standard voices cost $4.00, Neural $16.00, Generative $30 and Long-Form $100.00 per 1 million characters.
There is an important change for new accounts. AWS says that starting July 15, 2025, new customers receive up to $200 in AWS Free Tier credits that can be used on Polly and other services. You choose a free plan or a paid plan at sign-up, the free plan lasts 6 months, and all credits must be used within 12 months of creating the account. If you are opening a new AWS account today, plan around those credits rather than a fixed monthly character allowance.
ElevenLabs
ElevenLabs' Free plan includes 10k credits a month, and its API pricing page lists 10,000 characters included on the free tier for its v3 model. The voices are popular, but read the license terms before you ship anything. ElevenLabs' help center states that the free plan does not include a commercial license and cannot be used for any commercial purpose. If you publish audio made on the free plan, you must credit ElevenLabs by including "elevenlabs.io" or "11.ai" in the title. The first plan with a commercial license is Starter at $6 a month.
If you want ElevenLabs voices on a WordPress site, see ElevenLabs voices on WordPress (opens in a new tab). That setup uses your own ElevenLabs API key, so ElevenLabs bills you directly.
OpenAI text to speech
OpenAI's API pricing page does not list a free tier for its speech models, so you pay from the first request. As of September 2026 it lists tts-1 at $15.00 and tts-1-hd at $30.00 per 1 million characters. The newer gpt-4o-mini-tts model is priced per token: $0.60 per 1 million text input tokens and $12.00 per 1 million audio output tokens.
AtlasVoice API (our product)
We built the AtlasVoice API (opens in a new tab) for site owners and developers who want natural voices without cloud-console setup. It is a plain REST API: you send text and get back an MP3. The free plan includes 100,000 characters a month with no card needed, 81 languages and accents with the AtlasVoice TTS voice, and 30 requests a minute. Each request can carry up to 5,000 characters, so longer articles are sent in parts. Our terms allow you to use the audio you create for any lawful purpose, including commercial use.
Two things to know. Your text is never stored, and the generated MP3 is deleted from our servers within minutes, so save the file with your content. And keep your API key on your server, never in browser JavaScript. Here is a first request:
curl -X POST "https://api.atlasvoice.cloud/v1/synthesize?raw=1" \
-H "Authorization: Bearer $ATLASVOICE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"text": "Hello from AtlasVoice.", "lang": "en-US"}' \
-o hello.mp3If you outgrow the free allowance, Premium comes with an AtlasVoice Pro licence (from $59 a year) and has no monthly character limit. You can read more in our API announcement (opens in a new tab).
How to choose the best free text to speech API
There is no single best free text to speech API. The right one depends on what you are building:
- You need a large monthly allowance that does not expire. Google's 4 million monthly characters for Standard and WaveNet voices keeps renewing, as long as you are comfortable enabling billing. Polly's 5 million Standard characters is similar if you have an older AWS account.
- You are testing an idea with low traffic. Azure F0 or the AtlasVoice free plan let you build without a large commitment. Watch Azure's 20 transactions per minute limit.
- You are opening a new AWS account. Polly works, but budget with the $200 credits and the 6-month free plan in mind.
- You are publishing audio on a business website. Check commercial-use rules first. ElevenLabs' free plan cannot be used commercially, so read the terms of whichever API you pick.
- You want the least setup. A focused API with one key and one endpoint is quicker to try than a full cloud console. For other ways to add voice, including no-code options, see how to add text to speech to any website (opens in a new tab).
Also estimate your volume. A typical blog post is a few thousand characters, so a 100,000-character allowance covers a modest number of articles a month, while millions of characters cover a busy site. Since you pay per character, generate each article once, save the MP3, and replay the file instead of calling the API on every page view.
Is there a truly unlimited free text to speech API?
Not as a hosted server API. Every hosted text to speech API free of charge has a cap somewhere, because each request uses real computing power. There are two honest alternatives if you need "free and unlimited":
- The browser's Web Speech API. Modern browsers include
speechSynthesis, which reads text aloud using the device's own speech synthesizer. It costs nothing and has no character limit, but it runs in the visitor's browser, not on a server. Voices vary by device and browser, and the API has no built-in way to save the speech as an audio file. - Open-source engines you host yourself. Projects such as Piper (a fast, local neural TTS engine now developed at OHF-Voice/piper1-gpl under GPL-3.0) and Coqui TTS (MPL-2.0 code) run on your own hardware. There is no per-character bill, but you pay for servers and maintenance. Check each voice model's own license: Coqui's XTTS-v2 model, for example, is under the separate Coqui Public Model License.
For many sites, a mix works best: browser voices for instant, free playback, and a server API for natural voices saved as MP3.
Next step
If you want to try a free text to speech API today, create a free account at app.atlasvoice.cloud (opens in a new tab), copy your key, and run the curl example above. You will have your first MP3 in a few minutes, with 100,000 free characters a month to experiment with. See the full reference on the AtlasVoice API page (opens in a new tab).
