# 6 best text to speech APIs for insurance in 2026

By Samuel Oram-Jones

A first notice of loss call, where a policyholder reports a crash, a burst pipe, or a theft, comes from someone having a bad day, and a voice that reads its opening line like an announcement tells that caller nobody is listening. A voice agent that also handles renewals and billing has to shift tone between one call and the next.

Bland Speech leads because Bland Speech v3 ranked first of 13 AI voices in Design Arena's Audio Realism Bench results of 4 August 2026, and it takes inline tags such as [say warmly] and [sighs] that let a claims line sound calm on purpose. Each of the others steers tone, or clears a security review, in its own way, as of November 2026.

## What are the best text to speech APIs for insurance?

The best text to speech API for insurance in 2026 is Bland Speech, followed by Inworld, ElevenLabs, Google, Microsoft Azure, and Rime, with Bland Speech first on realism and phone audio and the others first on a specific way of steering tone or meeting a security review.

- **Bland Speech:** Best overall for claims and service calls where tone has to sound human
- **Inworld:** Best for teams that want to steer delivery in plain words
- **ElevenLabs:** Best for expressive audio tags and 90+ languages on its newest models
- **Google:** Best for teams that want Gemini-TTS style prompts under a Google Cloud BAA
- **Microsoft Azure:** Best for teams on a Microsoft agreement that want named speaking styles
- **Rime:** Best for regulated-industry voice agents that need HIPAA, SOC 2, and a BAA from one vendor

| API | How tone is steered | Compliance the company states | Phone audio (mu-law) | Audio Realism Bench, 4 August 2026 |
|---|---|---|---|---|
| Bland Speech | Inline tags, plus expressiveness and stability dials | Bland: SOC 2 Type II, HIPAA compliant with a signed BAA | Yes | #1 AI voice, 1365 |
| Inworld | Plain-language steering and 50 non-verbal sounds | SOC 2 Type II, HIPAA, BAA on Enterprise | Yes | TTS-1.5 Max: #13, 763 |
| ElevenLabs | Audio tags such as [whispering] and [laughing] | SOC 2 Type 2, HIPAA, BAAs for qualifying healthcare enterprises | Yes | Eleven v3: #12, 938 |
| Google | Style prompts and bracket tags on Gemini-TTS | Text-to-Speech under Google's BAA and SOC 2 scope | Yes | Gemini 2.5 Pro TTS: #5, 1109 |
| Microsoft Azure | Named styles with a degree setting | BAA through Microsoft's Product Terms | Yes | MAI-Voice-2: #2, 1224 |
| Rime | Pauses, speed, and a pronunciation lexicon | SOC 2 Type 2, HIPAA compliant, BAA available | Yes | Wasn't in those results |

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

- **How tone is steered:** the controls each company documents for delivery, such as inline tags, style prompts, or named styles.
- **Compliance the company states:** each company's own wording. HIPAA and BAA lines matter to insurers that handle health information.
- **Audio Realism Bench:** the rank among 13 AI voices and the score in Design Arena's published results.

</details>

Bland Speech takes the top spot for insurance because inline delivery tags and 8 kHz phone audio come with that 4 August 2026 realism result, and Inworld is the place to look for an insurer that would rather describe a delivery in a sentence than tag it.

## What should an insurer check in a text to speech API?

An insurer should check how much control the API gives over tone, which compliance statements the vendor makes for health lines, and whether audio comes back in the format phone lines use.

### How much control over tone does each API give?

Tone is the part of a speech API a claims team hears first. Bland Speech takes tags in square brackets, including [say warmly], [say quietly], [say sadly], [voice lowered], [sighs], and [exhales], on its current-generation voices, plus two dials from 0.0 to 1.0, expressiveness for how much intonation varies and stability for how much the same input varies between renders. Inworld's Realtime TTS-2 takes a delivery described in plain words, and ElevenLabs' Eleven v4 takes audio tags such as [whispering]. Google's Gemini-TTS takes a style prompt plus bracket tags and warns that spoken-aloud adjective tags are likely an undesired side effect, and Azure applies named styles with a degree from 0.01 to 2 on a subset of neural voices.

### Which compliance statements matter on a health line?

A health insurance line that passes member information to a speech vendor raises the same BAA question hospitals ask. Bland is HIPAA compliant with a signed BAA and SOC 2 Type II audited. Inworld, ElevenLabs, Rime, Google, and Microsoft each describe a BAA path, with Inworld offering it on Enterprise or as a Growth add-on and Google listing Text-to-Speech among the products its BAA covers.

### What phone audio does an insurance line need?

US phone networks carry mu-law, a compressed 8-bit format sampled at 8 kHz. All six APIs here return it, so phone audio separates none of them.

## How do the 6 best text to speech APIs for insurance compare?

These six differ most in how human they sound and in how a team tells the voice what to feel.

### 1. Bland Speech: Best overall for claims and service calls where tone has to sound human

**What does it do?** Bland Speech is the text to speech model inside Bland, a voice AI provider that runs its own speech model for enterprise phone calls. Text streams in over HTTP or over a WebSocket that takes text as a language model writes it, and audio comes back as PCM from 8 to 48 kHz or as mu-law at 8 kHz. In Bland's dashboard studio, a director feature called Norm writes and performs a script from a described scene. The [text to speech API](/text-to-speech-api) page has the endpoints.

<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 on Audio Realism Bench on 4 August 2026</li><li>Inline tags for warmth, volume, and sighs, plus timed pauses</li><li>Bland is HIPAA compliant with a signed BAA, and SOC 2 Type II audited</li><li>One standard rate, $0.04 per 1,000 characters, with no monthly plan</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, which is $40 per million, with 50,000 characters free for a new account. Bland puts the BAA under Enterprise. Enterprise pricing is custom, set with Bland's sales team. The [Bland Speech](/speech) page has a studio to try a voice.

### 2. Inworld: Best for teams that want to steer delivery in plain words

**What does it do?** Inworld sells realtime voice AI aimed at consumer applications, with Realtime TTS-2 as its flagship text to speech model and TTS-2 Flash as its fast one. TTS-2 takes a delivery described in plain words, as a request instruction or in brackets partway through a line, and both models render 50 non-verbal sounds such as [laugh] and [sigh]. It returns mu-law and A-law at 8 kHz and serves an OpenAI-compatible speech endpoint.

<div class="cmp-proscons" data-rival="inworld"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Delivery steered in plain words on TTS-2</li><li>SOC 2 Type II certified, and HIPAA compliant with a BAA on Enterprise or as a Growth add-on</li><li>Instant cloning from as little as 3 seconds of audio on every plan</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Steering works on TTS-2 only, and TTS-2 Flash ignores it</li><li>TTS-1.5 Max ranked 13th of 13 AI voices on Audio Realism Bench on 4 August 2026, and Realtime TTS-2 wasn't in those results</li><li>Zero data retention starts on the $300 a month Developer plan</li></ul></div></div>

**Pricing:** $25 per million characters for TTS-2 and $15 for TTS-2 Flash on pay as you go, falling to $12.50 and $7 on the $1,500 a month Growth plan, with Enterprise custom.

### 3. ElevenLabs: Best for expressive audio tags and 90+ languages on its newest models

**What does it do?** ElevenLabs sells text to speech, transcription, and voice agents. Eleven v4, its most emotionally rich model, and v4 Turbo launched on 28 September 2026, and audio tags such as [whispering], [shouting], and [laughing] work on v4. The API returns mu-law at 8 kHz.

<div class="cmp-proscons" data-rival="elevenlabs"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Audio tags on Eleven v4, with 90+ languages across the v4 family</li><li>SOC 2 Type 2 and HIPAA compliance, and BAAs for qualifying healthcare enterprises</li><li>Instant clones from 10 seconds of audio on v4</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Eleven v3 ranked 12th of 13 AI voices on Audio Realism Bench on 4 August 2026</li><li>Eleven v4 costs $0.08 per 1,000 characters</li><li>Eleven v4 and v3 don't take SSML break tags, so pauses come from punctuation and audio tags</li></ul></div></div>

**Pricing:** $0.08 per 1,000 characters for Eleven v4, v3, and Multilingual v2, and $0.04 for v4 Turbo, v3 Conversational, and Flash, against plans from Free to Business at $990 a month. [Bland Speech vs ElevenLabs](/compare/bland-speech-vs-elevenlabs) compares the two.

### 4. Google: Best for teams that want Gemini-TTS style prompts under a Google Cloud BAA

**What does it do?** Google Cloud Text-to-Speech offers Chirp 3: HD voices, Gemini-TTS models priced by token, and older WaveNet, Neural2, and Standard voices. Gemini-TTS takes a style prompt as the main driver of tone, plus bracket tags for sounds, styles, and pauses, and the API returns mu-law and A-law.

<div class="cmp-proscons" data-rival="google"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Gemini-TTS style prompts and bracket tags</li><li>Text-to-Speech is on the list of products Google's BAA covers</li><li>The first million Chirp 3: HD characters are free each month</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Chirp 3: HD has no emotion control, only speaking rate and pause tags</li><li>Vocalized-adjective tags can be spoken aloud, as Google warns</li></ul></div></div>

**Pricing:** $30 per million characters for Chirp 3: HD after the first free million each month, $4 for WaveNet and Standard, and Gemini 2.5 Pro TTS at $1.00 per million text tokens in and $20.00 per million audio tokens out. Gemini 2.5 Pro TTS ranked 5th of 13 AI voices on Audio Realism Bench on 4 August 2026.

### 5. Microsoft Azure: Best for teams on a Microsoft agreement that want named speaking styles

**What does it do?** Azure Speech offers neural voices in 100+ languages and locales, neural HD voices, custom voices, and Microsoft's MAI-Voice-2.1 models in public preview. A speaking style, a style degree, and a role are set at sentence level on a subset of neural voices, and HD voices pick up emotional cues from the text on their own.

<div class="cmp-proscons" data-rival="azure"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Named speaking styles with a degree from 0.01 to 2</li><li>A BAA through Microsoft's Product Terms, with no separate contract</li><li>Neural voices at $15 per million characters</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Styles work on a subset of neural voices, and some voices don't support every SSML tag</li><li>MAI-Voice is in public preview without a service-level agreement</li><li>Custom voices are limited to customers Microsoft manages directly</li></ul></div></div>

**Pricing:** $15 per million characters for Neural and Neural HD Flash voices, with 0.5 million characters free a month. MAI-Voice-2 ranked 2nd of 13 AI voices on Audio Realism Bench on 4 August 2026, and Azure Speech offers MAI-Voice-2.1 models today, with no public price shown for MAI-Voice.

### 6. Rime: Best for regulated-industry voice agents that need HIPAA, SOC 2, and a BAA from one vendor

**What does it do?** Rime sells conversational voice models, with Coda and Mist v3 for text to speech, and says they are built for voice agents, IVR, and telephony in healthcare, financial services, and other regulated industries. Rime documents custom pauses, speed control, and a pronunciation lexicon, and returns G.711 mu-law at 8 kHz.

<div class="cmp-proscons" data-rival="rime"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>HIPAA compliant since February 2024, per Rime, and SOC 2 Type 2 compliant</li><li>A BAA available, and dedicated VPC or on-premises deployment for enterprise customers</li><li>Word-level timestamps on Coda over WebSocket and gRPC</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Coda's cloud API serves 10 languages</li><li>Voice cloning is an enterprise order placed with Rime</li><li>Its voices weren't in Audio Realism Bench's results of 4 August 2026</li></ul></div></div>

**Pricing:** On the Flex plan, $0.030 a minute for Mist v3 and $0.050 a minute for Coda v2, and the Free plan includes $50 of usage every month. Rime says a minute is about 1,000 characters.

## Which text to speech sounds most like a real person on an insurance call?

In Design Arena's Audio Realism Bench results of 4 August 2026, Bland Speech v3 ranked first of 13 AI voices at 1365, behind only recorded human speech at 1500, ahead of Microsoft's MAI-Voice-2 at 1224, Google's Gemini 2.5 Pro TTS at 1109, and ElevenLabs' Eleven v3 at 938, with Inworld's TTS-1.5 Max 13th at 763. Design Arena plays listeners two unlabelled recordings of the same line and asks which sounds more like a real person, and a higher Elo rating means a voice was picked more often.

Rime and Inworld's Realtime TTS-2 weren't in those results, and ElevenLabs' Eleven v4 came out after them and isn't scored in them. Design Arena [published the method and scores](https://notes.designarena.ai/audio-realism-benchmark-measuring-realism-in-audio-models/) on 4 August 2026.

## How were these text to speech APIs compared?

Samuel Oram-Jones compared six APIs on tone control, compliance statements, phone audio, voice realism, and price per character. Every rival detail comes from that company's own pricing, documentation, and trust pages, and voice scores come from Design Arena's published results of 4 August 2026.

Bland's [comparison of text to speech APIs](/compare/text-to-speech) covers the same vendors by job, the [best lists](/compare/text-to-speech/best-lists) group them by use, and [how to choose a voice for a phone agent](/blog/choose-voice-for-phone-agent) covers the listening side.

## Frequently asked questions

### What is the best text to speech API for insurance?

Bland Speech is the best text to speech API for insurance calls in 2026 for teams that put how a voice sounds first. Bland Speech v3 ranked first of 13 AI voices on Audio Realism Bench on 4 August 2026, and it takes inline tags such as [say warmly] and [sighs].

### Which text to speech APIs let a team control tone?

Bland Speech takes bracket tags and two numeric dials, Inworld takes plain-language steering on TTS-2, and ElevenLabs takes audio tags on Eleven v4. Google's Gemini-TTS takes style prompts and bracket tags, and Azure takes named styles on a subset of neural voices.

### Does an insurer need a BAA for a text to speech API?

A health insurance line that passes member information to a speech vendor is a BAA question. Bland is HIPAA compliant with a signed BAA, and Inworld, ElevenLabs, Google, Microsoft, and Rime each describe a BAA path of their own.

### Can an insurer clone a claims adjuster's voice?

Bland Speech can clone a voice. An instant clone needs one clean sample of about 10 seconds, and a professional clone is fine-tuned on 30 minutes or more of audio and goes through Bland's consent and rights review. Every clone is private to the customer's organization, and the customer confirms the voice is theirs to use.

### How much does text to speech cost for insurance calls?

Among these six, list prices run from $4 per million characters for Google's WaveNet and Standard voices to $80 per million for ElevenLabs' Eleven v4, and Bland Speech costs $40 per million, billed as $0.04 per 1,000 characters with no monthly plan.
