Google, Microsoft and Amazon Text to Speech Compared for Small Teams

Google text to speech vs Amazon Polly vs Azure comes down to price tier and free allowance. Standard voices cost $4 per 1 million characters on Google and Amazon, and neural voices cost $15 to $16 on all three. At about 900 characters per minute of speech, that's roughly 1.4 to 1.5 cents a minute.

Last updated: October 14, 2026. Every price and free tier below was read from Google's, Amazon's and Microsoft's own pricing pages on that date.

This guide is for small teams choosing between the three big cloud speech services for training videos, product demos, phone menus or app audio. It shows real cost per minute, what you'd pay at 3 volumes, and the hidden cost that matters more than the per-character rate: the developer time to use an API.

Key Takeaways

• Neural voices cost about the same everywhere: Google Neural2 and Amazon Polly Neural are $16 per 1M characters, and Azure Neural is $15 (all retrieved 2026-10-14).
• Free tiers cover most small teams. Google gives 1M free Neural2 characters a month, Azure 0.5M, and Polly 1M for the first 12 months.
• Premium voices cost 1.5× to 10× more: Azure Neural HD is $22, Google Chirp 3 HD and Polly Generative are $30, Polly Long-Form is $100 and Google Studio is $160.
• The API is the real cost. None of the three has a script-to-MP3 editor, so you build or buy the interface.
• For most small teams, a no-code tool can win on total cost once developer time is counted.

On this page: Prices compared · Cost per minute · Monthly bills · Voices and features · No-code vs API · Decision table · Checklist · FAQ

Google text to speech vs Amazon Polly vs Azure: which is cheapest?

For standard voices, Google and Amazon tie at $4 per 1 million characters. For neural voices, Azure is cheapest at $15, with Google Neural2 and Polly Neural at $16. Premium tiers vary a lot, from $22 on Azure to $160 on Google Studio.

Prices below are pay-as-you-go list prices in USD, before any committed-use discount.

All prices are per 1 million characters sent for synthesis. Sources and retrieval dates are in the last section.

Two billing details matter. Google and Amazon charge for characters you send, not for the length of the audio. Amazon's pricing page also says you can cache and replay generated speech at no extra cost, which helps if you reuse prompts such as phone menus.

What does one minute of cloud text to speech cost?

At about 900 characters per minute of speech, standard voices cost about $0.0036 a minute and neural voices about $0.014. Premium voices run from $0.02 to $0.144 a minute. That's our calculation, using 150 words a minute and 6 characters per word including the space.

The Kveeky bar is a flat plan price divided by included minutes, so it isn't a like-for-like rate. It includes the editor, the voice library and MP3 or WAV export. The cloud rows are raw synthesis only.

Use the words to minutes calculator to turn your script length into minutes before you estimate characters.

For the cost of a whole video, including editing and revisions, see how much an AI voiceover costs per video.

What does a small team pay per month?

At 100 minutes a month, all three clouds cost between $0 and $1.44. At 5,000 minutes, they cost $56 to $72. Free allowances do most of the work at small volumes, so the per-character rate barely matters until you pass about 1,000 minutes.

*Polly Neural is shown after its 12-month free tier ends. Google and Azure free allowances reset monthly. Cloud figures are synthesis only; Kveeky figures are plan prices and include the editor.

Worked example. A team makes 20 training videos a month at 5 minutes each. That's 100 minutes, or about 90,000 characters. On Google Neural2 the bill is $0, because the 1M free allowance covers it. The same work on a Kveeky Basic yearly plan is $7.50 (about 100 minutes included), and that includes the editor and export.

The takeaway is plain: cloud TTS is the cheaper engine at every volume in this table. Whether it's the cheaper solution depends on who builds the tool around it.

How do the voices and features differ?

All three are developer services: you send text and get audio back. Google describes more than 220 voices across 40+ languages on its product page. Amazon Polly offers four engines (standard, neural, long-form and generative). Azure adds Neural HD and custom voices.

Google Cloud Text-to-Speech

• Best for: teams already on Google Cloud who want a large free allowance.
• Price: Standard $4, Neural2 $16, Chirp 3 HD $30, Studio $160 per 1M characters (retrieved 2026-10-14).
• Output: MP3, OGG_OPUS, and LINEAR16 with a WAV header, per Google's AudioEncoding reference.
• Pros: the biggest free allowances on neural tiers; clear tier names.
• Cons: Studio at $160 is the priciest tier here, and there's no editor for non-developers.

For the phone and browser features most people mean by "Google text to speech", see our guide to Google text to speech.

Amazon Polly

• Best for: teams on AWS, and apps that replay the same lines many times.
• Price: Standard $4, Neural $16, Generative $30, Long-Form $100 per 1M characters (retrieved 2026-10-14).
• Pros: the free tier includes 5M standard characters a month with no stated time limit, and caching generated speech costs nothing extra.
• Cons: the neural, long-form and generative free tiers last only 12 months, and Long-Form is expensive.

Microsoft Azure Speech

• Best for: teams on Microsoft 365 or Azure that want the cheapest neural rate.
• Price: Neural $15 and Neural HD $22 per 1M characters, per Microsoft's retail rate card (retrieved 2026-10-14).
• Pros: lowest neural price, and committed-use tiers drop the rate further at high volume.
• Cons: the smallest free allowance (0.5M neural characters a month), and the pricing page loads its figures with JavaScript, which makes it harder to audit.

Kveeky

• Best for: people who want to paste a script, pick a voice and download MP3 or WAV without writing code.
• Price: free plan with 500 credits a month (about 6.6 minutes, standard voices only, no credit card). Basic is $9/month, or $7.50/month billed yearly, for 100 minutes and all 700+ voices (kveeky.com/pricing.md, retrieved 2026-10-14).
• Pros: 700+ voices, 40+ languages with voices, voice, speed, volume and emotion controls, and commercial usage rights on every paid plan.
• Cons: it's a browser tool, not a cloud API. If you need to generate audio from your own app, use one of the three clouds above.

When is a no-code tool cheaper than a cloud API?

A no-code tool is cheaper when nobody on your team will build and maintain the integration. The break-even is developer time, not characters. Below, the figures marked "assumed" are illustrative; replace them with your own.

On those assumptions, the no-code route is cheaper by more than $1,000 a year at 100 minutes a month. The gap closes only when volume is high enough that per-character savings beat the labour, or when your product must generate audio automatically.

Pick the cloud API when audio is generated inside your app, when you need thousands of short clips on demand, or when you already have an engineer on that stack. Pick a no-code tool when people write scripts and download files, as with training videos, demos and social clips.

If you are weighing people against software, compare with the real cost of hiring a voice actor for 100 YouTube videos.

Which one should a small team choose?

Choose the cloud that matches your existing infrastructure, and choose a no-code tool if no engineer owns the audio. The three clouds are close enough on price that your current vendor and free allowance should decide.

If you're still comparing no-code options, see our best AI voice generators for small business and the free vs paid text to speech breakdown.

Checklist before you commit to a cloud API

Run through these 7 checks before you write any code. Each one catches a cost the price table hides.

Frequently asked questions

Is Google or Amazon text to speech cheaper?

They tie on standard voices at $4 per 1M characters and on neural voices at $16. Google's free allowances are larger and monthly, while Polly's neural free tier lasts only 12 months. Prices retrieved October 14, 2026.

Is Azure text to speech cheaper than Polly?

Slightly, for neural voices. Azure Neural is $15 per 1M characters against $16 for Polly Neural. Azure's free allowance is smaller, at 0.5M neural characters a month.

How much does one minute of cloud text to speech cost?

Neural voices cost about 1.4 to 1.5 cents a minute, assuming 900 characters per minute of speech. Standard voices cost about 0.4 cents, and premium tiers range from 2 cents to 14 cents.

Are cloud text to speech services free?

Each has a free allowance. Google gives 1M neural characters a month, Polly 1M for the first 12 months, and Azure 0.5M a month. Past that, billing starts automatically.

Do I need a developer to use Google, Amazon or Azure text to speech?

Yes, for practical use. They're APIs without a script editor, so someone has to write code or connect a tool. If you only need finished MP3 or WAV files, a no-code voice generator avoids that work.

More quick answers on cloud text-to-speech pricing

How we checked this guide

This guide is written by Ankit Agarwal for the Kveeky team. Disclosure: Kveeky makes an AI voice generator. We compared Kveeky's flat plan price with cloud per-character rates and state that the two aren't like-for-like. We didn't generate or listen to audio for this guide, so it makes no claims about voice quality.

• Google Cloud Text-to-Speech pricing: cloud.google.com/text-to-speech/pricing, retrieved 2026-10-14.
• Google audio formats: AudioEncoding reference, retrieved 2026-10-14.
• Amazon Polly pricing: aws.amazon.com/polly/pricing, retrieved 2026-10-14.
• Amazon Polly engines: SynthesizeSpeech API reference, retrieved 2026-10-14.
• Microsoft Azure Speech free tier: Azure Speech pricing, retrieved 2026-10-14. Paid rates (Neural $15, Neural HD $22) come from Microsoft's public Azure Retail Prices API, East US, USD, retrieved 2026-10-14.
• Kveeky plans and rights: kveeky.com/pricing.md, retrieved 2026-10-14.

Characters per minute (900) and the developer-time figures are our assumptions. Prices change, so confirm on the vendor's page before you budget.

Your next step: take one real script, count its characters, and price it in the table above. If the answer is a few dollars and nobody can build the tooling, test it free in the Kveeky voice generator for training videos.