# 6 Azure text to speech alternatives in 2026 for production voice agents

By Samuel Oram-Jones

Microsoft says its newest voices aren't ready for production. MAI-Voice-2.1 and MAI-Voice-2.1-Flash are in public preview, and Microsoft writes that "This preview is provided without a service-level agreement, and is not recommended for production workloads." Microsoft prices MAI-Voice at "N/A per 1M characters", as of November 2026, so a team has no published rate to budget from.

A preview label matters less in a demo than on a phone line that takes real calls. For that work Bland Speech comes first, because Bland Speech v3 ranked first of 13 AI voices in Design Arena's [Audio Realism Bench results of 4 August 2026](https://notes.designarena.ai/audio-realism-benchmark-measuring-realism-in-audio-models/) and Bland publishes its compliance on one trust page, with ElevenLabs, Cartesia, Google, Amazon Polly, and Rime after it.

## Why do teams look for Azure text to speech alternatives?

Teams look for Azure text to speech alternatives when the voices they want sit in preview, behind an access review, or without a posted price, because Microsoft's MAI-Voice models are in public preview, custom and personal voices are limited access, and Microsoft's price for MAI-Voice reads N/A.

Cloning on Azure runs through Microsoft's Limited Access process. Microsoft says custom neural voice is available by registration only and for certain use cases, and that "Only customers managed by Microsoft, meaning those who are working directly with Microsoft account teams, are eligible for access." Personal voice, which builds a voice from a 5 to 90 second sample plus a verbal consent statement, restricts its API to eligible customers and approved use cases, and MAI-Voice instant cloning is gated through the same review. Microsoft also requires explicit written permission from voice talent and reserves the right to use speaker recognition on the recorded consent statement.

Streaming text in is narrower than it first looks, because Azure takes text over its WebSocket v2 endpoint only through the Speech SDK in C#, C++, and Python.

## What should I check in an Azure text to speech alternative?

A buyer should check whether the voice is generally available or in preview, who may clone a voice and how, how text streams in, whether a rate is published for the voice that was auditioned, and what the compliance paperwork covers.

Production status comes first, and a service-level agreement is the vendor's written uptime commitment. Microsoft says MAI-Voice has none during preview, while ElevenLabs and Rime each include an SLA in their Enterprise plans.

Cloning comes second. Bland Speech needs about 10 seconds for an instant clone and 30 minutes or more for a professional one, ElevenLabs needs 10 seconds on a Starter plan and about 30 minutes on Creator or above, Cartesia needs 10 seconds on Pro and 30 minutes on Startup, and Rime sells voice cloning to Enterprise customers from 30 to 60 minutes of recordings.

Compliance paperwork comes third. Microsoft makes a HIPAA BAA available through its Product Terms by default to customers that are covered entities or business associates. Bland publishes SOC 2 Type II, HIPAA with a signed BAA, and FedRAMP 20x Class C at Moderate, granted on 17 November 2026.

## Which Azure text to speech alternative sounds most human?

Bland Speech v3 ranked first of 13 AI voices in Design Arena's Audio Realism Bench results of 4 August 2026, at 1365, and Microsoft's MAI-Voice-2 ranked 2nd at 1224, 141 points behind.

Azure Speech sells the newer MAI-Voice-2.1 and MAI-Voice-2.1-Flash in public preview, and neither was scored. Google's Gemini 2.5 Pro TTS ranked 5th at 1109, Cartesia's Sonic 3.5 ranked 6th at 1096, and ElevenLabs' Eleven v3 ranked 12th at 938. Amazon Polly and Rime weren't in those results. Cartesia's Sonic 3.6 launched on 27 August 2026 and ElevenLabs' Eleven v4 on 28 September 2026, after the results, so neither was scored in them.

Design Arena's listeners compare two recordings of the same text, blind, and pick the one that sounds more like a real person. The test used one female and one male American English voice per model, and prompts mostly ran up to about 40 words, so it speaks to short phone-style turns in American English.

## What are the best Azure text to speech alternatives?

The best Azure text to speech alternatives are Bland Speech, ElevenLabs, Cartesia, Google text to speech, Amazon Polly, and Rime, with Bland Speech first because it pairs the realism result of 4 August 2026 with a published rate per character and a clone from about 10 seconds of audio.

| Alternative | Best for | Published price | Audio Realism Bench, 4 August 2026 |
| --- | --- | --- | --- |
| Bland Speech | Voice agents on phone calls | $40 per million characters, pay as you go | Bland Speech v3: ranked 1st, 1365 |
| ElevenLabs | Self-serve cloning on a monthly plan | $40 to $80 per million characters, by model | Eleven v3: ranked 12th, 938 |
| Cartesia | A team that wants on-premise or own-VPC options | 1 credit per character, paid plans from $5 a month | Sonic 3.5: ranked 6th, 1096 |
| Google text to speech | A team already on Google Cloud | $30 per million characters for Chirp 3: HD | Gemini 2.5 Pro TTS: ranked 5th, 1109 |
| Amazon Polly | A team already on AWS | $4 to $100 per million characters, by engine | Wasn't in those results |
| Rime | Regulated phone lines with a per-minute rate | $0.050 a minute for Coda v2 on Flex | Wasn't in those results |

<details><summary>What these rows mean</summary>

- **Published price:** each vendor's own list rate with the model or plan named. Rime bills TTS by the minute and converts at about 1,000 characters a minute. Amazon Polly's four engines are Standard $4, Neural $16, Generative $30, and Long-Form $100.
- **Audio Realism Bench:** Design Arena's blind listening test of 4 August 2026, with rank among 13 AI voices and an Elo score, which is a rating built from head-to-head wins and losses. A vendor's newer models were not in those results.

</details>

### 1. Bland Speech

Bland Speech is a text to speech model that Bland runs itself, sold as a studio and an API. One endpoint streams over HTTP at POST /v2/tts, and a WebSocket at /v2/tts/ws takes text as a language model writes it. Output is mu-law at 8 kHz or PCM from 8 to 48 kHz, and Bland publishes 199ms to first audio. The [text to speech API](/text-to-speech-api) page has the request details.

Bland Speech suits a team that needs a clone without an access application: an instant clone needs one clean sample of about 10 seconds, a professional clone is fine-tuned on 30 minutes or more of audio after Bland's consent and rights review, and every clone is private to the customer's organization. Bland is SOC 2 Type II audited, HIPAA compliant with a signed BAA, and FedRAMP 20x Class C certified at Moderate, and Bland's trust page says speech synthesis runs inside the FedRAMP boundary. Enterprise customers get dedicated orchestration and dedicated GPUs run by Bland.

<div class="cmp-proscons cmp-proscons-own"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Bland Speech v3 ranked first of 13 AI voices in Audio Realism Bench results of 4 August 2026</li><li>Instant clones from about 10 seconds of audio, with no access application</li><li>Bland is SOC 2 Type II audited and HIPAA compliant with a signed BAA</li><li>Mu-law at 8 kHz and PCM from 8 to 48 kHz for phone lines</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>More than a hobby project needs, because Bland Speech is built for production calls</li></ul></div></div>

**Pricing:** $0.04 per 1,000 characters ($40 per million), pay as you go with no monthly plan, and 50,000 free characters once for a new account. Enterprise pricing is custom, set with Bland's sales team. The [Bland Speech](/speech) page has a studio.

### 2. ElevenLabs

ElevenLabs sells text to speech, speech to text, and voice agents, and launched Eleven v4 and v4 Turbo on 28 September 2026. ElevenLabs suits a team that wants 90+ languages and a published ladder of plans: Starter is $6 a month, Creator $22, Pro $99, Scale $299, and Business $990. ElevenLabs says text to speech can run inside a customer's own cloud infrastructure, and BAAs for HIPAA customers sit on Enterprise.

<div class="cmp-proscons" data-rival="elevenlabs"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Instant cloning from 10 seconds of audio on Eleven v4</li><li>Median latency of about 100ms for Eleven v4 Turbo, excluding application and network latency</li><li>Mu-law and A-law output at 8 kHz</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Eleven v4 bills $0.08 per 1,000 characters, twice Eleven v4 Turbo's $0.04</li><li>Data residency is an Enterprise feature</li><li>The Free plan is non-commercial</li></ul></div></div>

**Pricing:** API usage is $0.08 per 1,000 characters for Eleven v4 and $0.04 for Eleven v4 Turbo and Flash, with Enterprise custom.

### 3. Cartesia

Cartesia sells Sonic 3.6 in 44 languages and says first audio arrives in under 90ms. Cartesia suits a team with deployment requirements, because Cartesia says Sonic 3.6 can be deployed on-premise inside a data center, including air-gapped environments, in a team's own VPC on AWS, GCP, or Azure, or through OEM licensing, under enterprise contracts. BAAs and DPAs sit on Enterprise.

<div class="cmp-proscons" data-rival="cartesia"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>On-premise, air-gapped, and own-VPC deployment under enterprise contracts</li><li>Instant clones from 10 seconds on Pro, and free professional clone training</li><li>SOC 2 Type II certified</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Credits and overage settings to manage each month</li><li>Emotion controls are in beta and work only in English</li></ul></div></div>

**Pricing:** Free is $0 with 20,000 credits, Pro is $5 with 100,000, Startup is $49 with 1.25 million, and Scale is $299 with 8 million, at one credit per character. Enterprise is custom.

### 4. Google text to speech

Google Cloud sells Chirp 3: HD voices at $30 per million characters and Gemini-TTS models priced per token. Google text to speech suits a team already on Google Cloud, because Text-to-Speech appears in Google's HIPAA covered-products list and in its SOC 2 scope, and Google states 380+ voices across 75+ languages and variants.

<div class="cmp-proscons" data-rival="google"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Chirp 3: HD includes the first million characters free each month</li><li>MULAW and ALAW encodings for phone audio</li><li>New customers get up to $300 in free Google Cloud credits</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Chirp 3: Instant Custom Voice no longer takes new access requests, and Gemini 3.8 voice replication is in Preview</li><li>Gemini-TTS has no free usage</li></ul></div></div>

**Pricing:** Chirp 3: HD is $30 per million characters, WaveNet and Standard voices are $4 per million, Neural2 is $16, and Studio is $160.

### 5. Amazon Polly

Amazon Polly sells four voice engines: Standard, Neural, Long-Form, and Generative. Amazon Polly suits a team already on AWS that reads out alerts at volume, because Standard voices cost $4 per million characters and the free tier covers 5 million Standard characters a month. Polly is a HIPAA Eligible Service under the AWS BAA and in scope for SOC 1, 2, and 3, and output includes mu-law and A-law at 8 kHz.

<div class="cmp-proscons" data-rival="amazon-polly"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Standard voices at $4 per million characters</li><li>HIPAA Eligible Service under the AWS BAA</li><li>Generated speech can be cached and replayed at no extra cost</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Custom voices go through an AWS account manager</li><li>AWS says speech marks aren't available for Generative voices</li></ul></div></div>

**Pricing:** Standard is $4 per million characters, Neural $16, Generative $30, and Long-Form $100, with a Standard free tier of 5 million characters a month.

### 6. Rime

Rime sells conversational voice models for enterprise voice agents, IVR, and telephony. Coda is the flagship text to speech model, Mist v3 is the low-latency model, and Rime is SOC 2 Type 2 and HIPAA compliant, with BAAs available. Rime suits a regulated phone line that wants a per-minute rate, and Rime says enterprise customers can run Rime in a dedicated VPC or fully on-premises.

<div class="cmp-proscons" data-rival="rime"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>G.711 mu-law output with 8 kHz sampling for telephony</li><li>$50 of usage free every month on the Free plan</li><li>HIPAA compliant, with BAAs and a SOC 2 Type II report on Enterprise</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Voice cloning is an enterprise order trained on 30 to 60 minutes of recordings</li><li>Coda serves 10 cloud API languages, and each voice serves one</li></ul></div></div>

**Pricing:** Flex has no monthly fee, with Mist v3 at $0.030 a minute and Coda v2 at $0.050. Scale is $100 a month at 20% off every rate, and Enterprise is custom.

## Which Azure text to speech alternative should I pick?

Azure's catch for a production voice agent is that its newest voice is a preview and its custom voice is an application. Bland Speech answers the second directly, since an instant clone starts from about 10 seconds of audio with no access review, and Bland prices v3 at a published $0.04 per 1,000 characters. With the realism result of 4 August 2026 on top, Bland Speech is the Azure text to speech alternative for a voice agent on phone calls.

Azure keeps real advantages. Azure's Neural voices cost $15 per million characters against Bland Speech's $40, commitment tiers cut that to as low as $7.50, standard voices cover 100+ languages and locales, and Azure Speech containers let a team host the service on its own hardware. A team on Azure's cloud that needs no cloned voice and has no realism bar for callers has less reason to leave.

[Bland Speech vs Azure text to speech](/compare/bland-speech-vs-azure-tts) and the [Azure text to speech review](/blog/azure-text-to-speech-review) go deeper on Azure, [Azure text to speech vs Amazon Polly](/blog/azure-text-to-speech-vs-amazon-polly) sets the two clouds side by side, and the [text to speech hub](/compare/text-to-speech) and the [alternatives family page](/compare/text-to-speech/alternatives) pick up providers beyond these six.

## Frequently asked questions

### What is the best Azure text to speech alternative?

Bland Speech is the best Azure text to speech alternative for production voice agents, because Bland Speech v3 ranked first of 13 AI voices in Design Arena's Audio Realism Bench results of 4 August 2026, a rate is published at $0.04 per 1,000 characters, and an instant clone needs no access application.

### Is Azure MAI-Voice ready for production?

Microsoft says it isn't: the MAI-Voice preview is provided without a service-level agreement and is not recommended for production workloads, and Microsoft prices MAI-Voice at N/A.

### Which Azure text to speech alternative lets a team clone a voice without applying for access?

Bland Speech, ElevenLabs, and Cartesia each start a clone from a self-serve account. Bland Speech needs one clean sample of about 10 seconds and the customer's confirmation that the voice is theirs to use, ElevenLabs needs a Starter plan, and Cartesia needs Pro. Azure's custom and personal voices are limited access.

### What compliance does Bland publish?

Bland is SOC 2 Type II audited, HIPAA compliant with a signed BAA, and FedRAMP 20x Class C certified at Moderate, granted on 17 November 2026. Bland's trust page says speech synthesis runs inside the FedRAMP boundary.

### How does Azure's price compare with Bland Speech's?

Azure's Neural voices cost $15 per million characters and Bland Speech costs $40 per million, so at 10 million characters a month the list prices are $150 and $400 before Azure's free allowance of 0.5 million characters and its commitment tiers. Microsoft charges $22 per million characters for Neural HD in East US, and MAI-Voice shows N/A.

### Can a team migrate from Azure text to speech to Bland Speech?

Yes, and the [migration guide](/blog/migrate-from-azure-tts) walks through the move. A voice cloned on Azure doesn't move, so a team re-clones from the original recordings with the speaker's consent.
