# 6 best text to speech APIs for sales calls in 2026

By Samuel Oram-Jones

A sales call is the one phone call where the listener owes the voice nothing. A caller with a billing problem stays on the line because the problem is theirs, while a prospect who hears a recording reading a script has every reason to hang up before the second sentence.

That makes delivery the first thing a buyer should test in a text to speech API, ahead of price. Bland Speech v3 ranked first of 13 AI voices in Design Arena's Audio Realism Bench results of 4 August 2026, and the five other APIs here lead on a lower list price, a faster claimed first audio, or a better fit with a cloud a team already pays for, as of November 2026.

## What are the best text to speech APIs for sales calls?

The best text to speech API for sales calls in 2026 is Bland Speech, followed by ElevenLabs, Grok TTS, Cartesia, Inworld, and Google, with Bland Speech first because its voice ranked first among AI voices in Design Arena's blind listening results of 4 August 2026 and it takes delivery tags inside the text.

- **Bland Speech:** Best overall for outbound sales calls that should sound like a person
- **ElevenLabs:** Best for audio tags on Eleven v4 and a rep's clone from 10 seconds of audio
- **Grok TTS:** Best for a low list price with inline and wrapping delivery tags
- **Cartesia:** Best for a first-audio claim under 90ms
- **Inworld:** Best for steering delivery with a plain-words description
- **Google:** Best for teams on Google Cloud that want Gemini-TTS tags

| API | How delivery is shaped | Price as the company publishes it | Audio Realism Bench, 4 August 2026 |
|---|---|---|---|
| Bland Speech | Tags such as [laughs] and [say warmly], plus expressiveness and stability controls | $0.04 per 1,000 characters | #1 AI voice, 1365 |
| ElevenLabs | Audio tags on Eleven v4 | $0.04 to $0.08 per 1,000 characters, by model | Eleven v3: #12, 938 |
| Grok TTS | Inline tags and wrapping tags | $15 per million characters | #3, 1156 |
| Cartesia | Emotion tags in beta for English, plus speed and volume | 1 credit per character on monthly plans | Sonic 3.5: #6, 1096 |
| Inworld | Plain-words steering on TTS-2, plus 50 non-verbal sounds | $25 per million characters for TTS-2, on demand | TTS-1.5 Max: #13, 763 |
| Google | Gemini-TTS bracket tags and a style prompt | $30 per million characters for Chirp 3: HD, Gemini-TTS by token | Gemini 2.5 Pro TTS: #5, 1109 |

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

- **How delivery is shaped:** the controls each company documents for tone, pace, and sound effects inside a line.
- **Price:** each company's list price with the model named.
- **Audio Realism Bench:** the rank among 13 AI voices and the score in Design Arena's published results.

</details>

A sales call has to be believed before anything said on it matters, and the blind listening result of 4 August 2026 is the one measure here that tests that directly, which puts Bland Speech first.

## What makes a text to speech voice work on a sales call?

A voice works on a sales call when it can change tone inside a sentence and stop when a prospect cuts in.

### Can the voice laugh, warm up, and slow down?

Bland Speech takes tags in square brackets right before the words they shape, from emotions such as [say warmly] and [say confidently] to vocalizations such as [chuckles] and [sighs]. Two controls from 0.0 to 1.0 sit beside them, expressiveness for how much intonation varies and stability for how much the same input varies between renders, and Bland recommends raising expressiveness for marketing reads and lowering it for disclosures and confirmations, which is the split a sales call has, a pitch followed by a price.

The other five shape delivery in their own ways. ElevenLabs runs audio tags such as [whispering] and [shouting] on Eleven v4, Grok TTS offers inline tags and wrapping tags such as <soft> and <loud>, Cartesia adds emotion tags in beta for English, and Inworld's Realtime TTS-2 takes a plain-words description of the delivery. Google's Gemini-TTS takes bracket tags and a style prompt, and Google warns that tags naming a feeling, such as [scared], can be spoken aloud, which it calls "likely an undesired side effect for most use cases".

### What happens when a prospect cuts in?

On Bland's WebSocket, a new context_id preempts the turn that is playing, which Bland says makes "interruption and barge-in a normal part of the protocol", and a buyer should confirm the same cancel path in any other API.

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

These six differ most in how human they sound and how much delivery control sits inside the text.

### 1. Bland Speech: Best overall for outbound sales calls that should sound like a person

**What does it do?** Bland Speech is the text to speech model inside Bland, a voice AI provider that runs its own model and speech. 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.

**Who is it for?** Bland Speech suits teams putting a voice on sales lines where the voice has to change register between a pitch and a price. The studio includes Norm, a director that writes and performs a script from a described scene.

<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>Tags such as [laughs] and [say warmly], plus expressiveness and stability controls</li><li>A new context on the WebSocket preempts the turn that is playing</li><li>Bland publishes 199ms to first audio</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>More than a one-off narration 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, pay as you go with no monthly plan, and 50,000 characters free once for a new account. Enterprise pricing is custom, set with Bland's sales team. The [Bland Speech](/speech) page has a studio to try a voice, and the [text to speech API](/text-to-speech-api) page has the endpoints.

### 2. ElevenLabs: Best for audio tags on Eleven v4 and a rep's clone from 10 seconds of audio

**What does it do?** ElevenLabs sells text to speech, transcription, and voice agents. Eleven v4 and v4 Turbo launched on 28 September 2026, v4 takes audio tags, and audio returns as mu-law at 8 kHz.

**Who is it for?** ElevenLabs suits sales teams that want clones of individual reps from a monthly plan.

<div class="cmp-proscons" data-rival="elevenlabs"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Audio tags on Eleven v4</li><li>Instant clones from 10 seconds of audio on v4, from the Starter plan</li><li>90+ languages on the Eleven v4 family</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, its most expressive model, costs $0.08 per 1,000 characters</li><li>The standard text to speech WebSocket doesn't accept Eleven v3 or v4</li></ul></div></div>

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

### 3. Grok TTS: Best for a low list price with inline and wrapping delivery tags

**What does it do?** Grok TTS is a text to speech API from xAI, a company now named SpaceXSI. It streams over HTTP and a WebSocket, and it returns mu-law and A-law at 8 kHz.

**Who is it for?** Grok TTS suits teams that want expressive tags at $15 per million characters with no plan tiers.

<div class="cmp-proscons" data-rival="grok-tts"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>$15 per million characters, with no plan tiers or monthly fee</li><li>Inline tags such as [laugh] and [sigh], and wrapping tags such as <soft> and <loud></li><li>Ranked 3rd of 13 AI voices on Audio Realism Bench on 4 August 2026</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Cloning through the API needs an Enterprise plan, with up to 30 voices free in the console</li><li>Custom voices are available in the United States, excluding Illinois</li></ul></div></div>

**Pricing:** $15.00 per 1 million characters, pay as you go.

### 4. Cartesia: Best for a first-audio claim under 90ms

**What does it do?** Cartesia sells Sonic, a text to speech model. Sonic 3.6, released on 27 August 2026, speaks 44 languages and streams over HTTP, server-sent events, and a WebSocket. Audio returns as mu-law or A-law at 8 kHz.

**Who is it for?** Cartesia suits teams that want speed and volume controls and one model for several markets.

<div class="cmp-proscons" data-rival="cartesia"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Cartesia says first audio arrives in under 90ms</li><li>Speed and volume controls, plus emotion tags</li><li>Instant voice clones from 10 seconds of audio</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Sonic 3.5 ranked 6th of 13 AI voices on Audio Realism Bench on 4 August 2026</li><li>Emotion tags are in beta and work in English only</li><li>Each break tag counts as 1 credit</li></ul></div></div>

**Pricing:** 1 credit per character from monthly plans: Pro at $5 for 100,000 credits, Startup at $49 for 1.25 million, and Scale at $299 for 8 million, with overages at $65, $45, or $38 per million.

### 5. Inworld: Best for steering delivery with a plain-words description

**What does it do?** Inworld sells realtime voice AI. Realtime TTS-2, launched on 5 May 2026, takes a plain-words description of delivery, and both TTS-2 and its Flash model render 50 non-verbal sounds such as [laugh] and [sigh]. Audio returns as mu-law or A-law at 8 kHz.

**Who is it for?** Inworld suits teams that would rather describe a delivery in a sentence than learn a tag list.

<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 Realtime TTS-2</li><li>Instant cloning from as little as 3 seconds of audio on every plan</li><li>On-Demand plan at $0 a month with a commercial license</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><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>Steering is supported only on TTS-2, not on Flash</li></ul></div></div>

**Pricing:** On-Demand bills $25 per million characters for TTS-2 and $15 for TTS-2 Flash with no monthly fee, and paid plans from $25 a month lower the rate.

### 6. Google: Best for teams on Google Cloud that want Gemini-TTS tags

**What does it do?** Google Cloud Text-to-Speech offers Chirp 3: HD voices, Gemini-TTS models priced by token, and older WaveNet and Neural2 voices. Gemini-TTS takes bracket tags for sounds such as [sigh], styles such as [shouting], and pacing such as [short pause], and a style prompt sets the overall delivery. Audio returns as mu-law.

**Who is it for?** Google suits sales teams already building on Google Cloud that want Gemini-TTS tags under an agreement they already hold.

<div class="cmp-proscons" data-rival="google"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Gemini-TTS takes bracket tags and a style prompt</li><li>Gemini 2.5 Pro TTS ranked 5th of 13 AI voices on Audio Realism Bench on 4 August 2026</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>Tags that name a feeling can be spoken aloud, which Google calls an undesired side effect for most use cases</li><li>Gemini-TTS is billed per audio token from the first token</li></ul></div></div>

**Pricing:** $30 per million characters for Chirp 3: HD after the first million each month, and $1.00 per million text tokens in plus $20.00 per million audio tokens out for Gemini 2.5 Pro TTS.

## Which text to speech sounds most like a person on a sales call?

Bland Speech v3 scored 1365 in the blind tests Design Arena published on 4 August 2026, and only a human recording scored higher, at 1500. Microsoft's MAI-Voice-2 ranked second at 1224, xAI's Grok TTS third at 1156, Google's Gemini 2.5 Pro TTS fifth at 1109, Cartesia's Sonic 3.5 sixth at 1096, ElevenLabs' Eleven v3 twelfth at 938, and Inworld's TTS-1.5 Max thirteenth at 763.

The test is narrower than a sales call. In [Design Arena's Audio Realism Bench](https://notes.designarena.ai/audio-realism-benchmark-measuring-realism-in-audio-models/), vetted native speakers hear two clips of the same line and pick the one that sounds more like a real person, and the Phone Agents category holds single turns that tend to be short. Each model was represented by one female and one male voice chosen as its most conversational American English voice, so the scores say nothing about how a cloned voice sounds. Cartesia's Sonic 3.6 and ElevenLabs' Eleven v4 came out after those results and weren't scored in them.

## How were these text to speech APIs compared?

The six were compared on voice realism, delivery controls, interruption handling, phone audio, and price per character. Each rival detail comes from that company's own pricing and documentation, each latency figure is the company's own claim, and voice scores come from Design Arena's results of 4 August 2026. Microsoft's MAI-Voice models are left out because Microsoft offers them in public preview without a service-level agreement and recommends against production workloads.

Bland's [comparison of text to speech APIs](/compare/text-to-speech) covers the same companies by job, and the [best lists](/compare/text-to-speech/best-lists) group sits beside it. [Best text to speech for outbound calls](/blog/best-tts-for-outbound-calls) covers dialing out.

## Frequently asked questions

### What is the best text to speech API for sales calls?

Bland Speech is the best text to speech API for sales calls in 2026. Bland Speech v3 ranked first of 13 AI voices on Audio Realism Bench on 4 August 2026, it takes delivery tags such as [laughs] and [say warmly] inside the text, and it returns mu-law audio for phone lines.

### Can text to speech laugh or change tone on a sales call?

Yes, on every API here. Bland Speech takes bracketed tags, ElevenLabs takes audio tags on Eleven v4, Grok TTS takes inline and wrapping tags, Cartesia takes emotion tags in English, Inworld takes a plain-words description on TTS-2, and Google's Gemini-TTS takes bracket tags.

### Which of these costs the least per character?

Grok TTS and Inworld's TTS-2 Flash model cost $15 per million characters, the lowest list prices here, and Google's Chirp 3: HD costs $30. Bland Speech costs $40 per million, and ElevenLabs charges $40 to $80 by model.

### Which text to speech API is fastest?

Each company publishes its own figure. Cartesia says first audio arrives in under 90ms, ElevenLabs states about 100ms at the median for v4 Turbo excluding network time, Inworld says as low as 50ms for its Flash model, and Bland publishes 199ms to first audio.

### Do these APIs return audio that phone lines accept?

Yes. All six return mu-law, the compressed format US phone networks carry, so none of them needs a conversion step on a call.

### Can a sales rep's voice be cloned?

Yes. Bland Speech makes an instant clone from one clean sample of about 10 seconds, every clone is private to the customer's organization, and the customer confirms the voice is theirs to use. ElevenLabs and Cartesia also start at 10 seconds, and Inworld says as little as 3 seconds. [Best voice cloning for sales calls](/blog/best-voice-cloning-for-sales-calls) compares every API that clones from a short sample.
