# 6 best text to speech APIs for outbound calling in 2026

By Samuel Oram-Jones

The person who picks up an outbound call never asked for it, so an agent's first sentence has to sound like a person before it says anything useful. A voice that sounds read out gets judged on that opening line, and a campaign says the same opening line tens of thousands of times.

Dialing is also lumpy, 50,000 calls in one week and none for the next month, and a plan with a monthly character allowance charges for the quiet weeks. Five of the six APIs here bill by use with no monthly fee, and ElevenLabs bills inside monthly plans.

## What are the best text to speech APIs for outbound calling?

The best text to speech API for outbound calling as of November 2026 is Bland Speech, because Bland Speech v3 ranked first of 13 AI voices in Design Arena's Audio Realism Bench results of 4 August 2026, it is billed per character with no monthly plan, and its WebSocket takes text as a language model writes it.

- **Bland Speech:** Best overall for outbound calls that open with a voice that sounds like a person
- **Grok TTS:** Best for the lowest pay as you go rate here, $15 per million characters
- **Deepgram:** Best for $200 of free credit and a self-hosting option
- **Rime:** Best for per-minute billing from a company that builds for telephony
- **Smallest AI:** Best for a low per-character rate with SIP trunking and phone numbers in the same company
- **ElevenLabs:** Best for expressive audio tags on Eleven v4

| API | Billing | Cost for 50,000 one-minute calls | First audio, as published | Audio Realism Bench, 4 August 2026 |
|---|---|---|---|---|
| Bland Speech | $0.04 per 1,000 characters, pay as you go | $1,600 | 199ms | #1 AI voice, 1365 |
| Grok TTS | $15 per million characters, pay as you go | $600 | Sub-second, for its voice APIs | #3, 1156 |
| Deepgram | $45 per million on Flux TTS, pay as you go | $1,800 | As low as 80ms | Wasn't in those results |
| Rime | $0.05 a minute on Coda v2 and $0.03 on Mist v3, no monthly fee on Flex | $2,500 or $1,500 | 30ms P50 on Coda v2, on the server | Wasn't in those results |
| Smallest AI | $19.50 per million on Lightning v3.1 Pro, pay as you go | $780 | Sub-100ms | Lightning v3.1 Pro: #11, 952 |
| ElevenLabs | $40 per million on v4 Turbo, inside monthly plans | $1,600 at that rate | About 100ms median on v4 Turbo | Eleven v3: #12, 938 |

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

- **Cost for 50,000 one-minute calls:** 50,000 minutes of agent speech at 800 characters a minute, which is 40 million characters (Cartesia converts a minute to 750 to 800 credits, Gradium to about 750 characters). Rime bills by the minute, so its cells are 50,000 minutes at each rate. ElevenLabs includes characters in a monthly fee, so its cell applies the per-character rate and leaves out plan fees.
- **First audio, as published:** each company's own figure, measured on its own terms, and not a race Bland ran.
- **Audio Realism Bench:** the model that was scored, its place among the 13 AI voices, and its score. Design Arena's human recordings scored 1500.

</details>

Bland Speech comes first because outbound calls open on a voice the listener has never heard, and Bland Speech v3 ranked first of 13 AI voices on 4 August 2026 for sounding like a real person.

## How were these compared, and what should an outbound team check?

Each API is compared on how the voice sounds, how it bills when a campaign starts and stops, how fast audio starts as each company publishes it, and whether a team can clone a sales rep's voice.

**Billing shape.** Bland Speech, Grok TTS, Deepgram, and Smallest AI bill per character as used, Rime bills per minute, and ElevenLabs includes characters in a monthly plan. Bland Speech's credits do not expire.

**First audio.** Time to first audio is the gap between sending text and hearing the first sound. Deepgram says as low as 80ms, Smallest AI sub-100ms, ElevenLabs a median of about 100ms on v4 Turbo before application and network time, Rime 30ms at the server, and Bland publishes 199ms, the slowest published figure here. A shorter number buys a shorter silence after a person says hello, and the voice that ranked first on 4 August 2026 buys what the person hears next.

**Phone audio.** All six return mu-law at 8 kHz, the compression US phone networks use, sampled 8,000 times a second.

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

These six differ most in how a campaign pays for them and how a voice sounds to a stranger, and what sets Bland Speech apart is the voice that ranked first on realism on 4 August 2026.

### 1. Bland Speech: Best overall for outbound calls that open with a voice that sounds like a person

**What does it do?** Bland Speech is the text to speech model inside Bland, which runs its own speech model instead of wrapping a third party's. Bland Speech streams over HTTP at POST /v2/tts or over a WebSocket at /v2/tts/ws that takes text as a language model writes it, and a new context interrupts the turn that is playing when the person on the line talks over the agent. Tags such as [say warmly] and [laughs] shape delivery, and the [text to speech API](/text-to-speech-api) returns mu-law at 8 kHz.

**Who is it for?** Bland Speech suits a team running outbound campaigns that wants the first sentence to sound like a person, with billing that follows use.

<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>One standard rate of $0.04 per 1,000 characters, with no monthly plan and credits that do not expire</li><li>A WebSocket that takes language model text as it is written and stops the playing turn on interruption</li><li>Mu-law and PCM at 8 kHz, ready for a phone line</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 once for a new account. Enterprise pricing is custom, set with Bland's sales team. A campaign's opening line can be tried in the [Bland Speech](/speech) studio before a single call goes out.

### 2. Grok TTS: Best for the lowest pay as you go rate here, $15 per million characters

**What does it do?** Grok TTS is the text to speech API from xAI, now named SpaceXSI. Grok TTS streams over a POST /v1/tts endpoint and a bidirectional WebSocket that takes text as deltas, returns mp3, wav, pcm, mulaw, and alaw, and offers 8 kHz as its narrowband telephony rate. Inline speech tags such as [laugh], [sigh], and [pause] shape delivery.

**Who is it for?** Grok TTS suits a team that wants the lowest per-character rate here from a voice that ranked 3rd of 13.

<div class="cmp-proscons" data-rival="grok-tts"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>$15 per million characters, pay as you go, with no plan tiers</li><li>Grok TTS ranked 3rd of 13 AI voices on Audio Realism Bench on 4 August 2026</li><li>An invisible watermark on every audio file by default, including cloned voices</li><li>Cloning from a reference clip in the console, 30 custom voices free</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Custom voices through the API are limited to Enterprise teams</li><li>Custom voices are available in the United States only, excluding Illinois</li></ul></div></div>

**Pricing:** $15.00 per million characters, billed as used.

### 3. Deepgram: Best for $200 of free credit and a self-hosting option

**What does it do?** Deepgram sells speech recognition and text to speech. Flux TTS streams over a WebSocket at /v2/speak and starts, Deepgram says, in as low as 80ms at 24 kHz. Mu-law and A-law output default to 8 kHz, and Flux TTS has a delivery dial from calm to animated, marked beta.

**Who is it for?** Deepgram suits a team that wants to test on free credit and may later run speech in its own cloud.

<div class="cmp-proscons" data-rival="deepgram"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Deepgram says Flux TTS starts in as low as 80ms</li><li>$200 of free credit, with no card and no expiry</li><li>Self-hosted and private cloud deployment for Flux TTS</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Deepgram's voices weren't in Audio Realism Bench's results of 4 August 2026</li><li>Flux TTS costs $0.045 per 1,000 characters on pay as you go</li></ul></div></div>

**Pricing:** $0.045 per 1,000 characters for Flux TTS, $0.030 for Aura-2, and $0.015 for Aura-1 on pay as you go, with Growth from $4,000 a year prepaid.

### 4. Rime: Best for per-minute billing from a company that builds for telephony

**What does it do?** Rime sells Coda and Mist, text to speech models for production voice agents, IVR, and telephony, and Prism, a speech-to-speech model. Coda, released in May 2026, streams over HTTP, WebSocket, and gRPC. Audio comes back as G.711 mu-law, with a sampling rate of 8000 set for a telephony stream.

**Who is it for?** Rime suits a team in healthcare or financial services that wants billing by the minute and an engine that can run in a dedicated VPC or on-premises.

<div class="cmp-proscons" data-rival="rime"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>No monthly fee on Flex, and $50 of free usage a month</li><li>HIPAA compliant since February 2024, SOC 2 Type 2, and a BAA available</li><li>30ms to first audio for Coda v2 at the server, as Rime publishes it</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Rime's voices weren't in Audio Realism Bench's results of 4 August 2026</li><li>Voice cloning is an Enterprise order that needs 30 to 60 minutes of recordings</li><li>Coda v2 costs $0.05 a minute, about 1,000 characters a minute</li></ul></div></div>

**Pricing:** On Flex, Mist v3 costs $0.030 a minute and Coda v2 $0.050, and Scale at $100 a month takes 20% off every rate. Rime's own benchmark puts Coda at 96ms for one request and 150ms with 12 at once.

### 5. Smallest AI: Best for a low per-character rate with SIP trunking and phone numbers in the same company

**What does it do?** Smallest AI sells Lightning text to speech, Pulse speech to text, and a voice agent platform with SIP trunking and managed phone numbers. Lightning v3.1 Pro streams over HTTP, server-sent events, and a WebSocket. Output is PCM, WAV, MP3, or mu-law from 8 kHz up.

**Who is it for?** Smallest AI suits a team that wants a low per-character rate and dialing and voice from one vendor.

<div class="cmp-proscons" data-rival="smallest-ai"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>$19.50 per million characters on Lightning v3.1 Pro, pay as you go</li><li>$10 in free credits to start</li><li>Smallest AI says first audio arrives in under 100ms</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Lightning v3.1 Pro ranked 11th of 13 AI voices on Audio Realism Bench on 4 August 2026</li><li>Lightning v3.1 retires on 30 November 2026, and voice clones made on it are affected</li></ul></div></div>

**Pricing:** About $0.195 per 10,000 characters for Lightning v3.1 Pro, which is $19.50 per million, and about $17.50 per million for Lightning v3.1.

### 6. ElevenLabs: Best for expressive audio tags on Eleven v4

**What does it do?** ElevenLabs sells text to speech, transcription, and voice agents. Eleven v4 Turbo, launched on 28 September 2026 as its real-time model, has a median latency of about 100ms before application and network time, and audio tags such as [whispering] and [laughing] shape delivery. The API returns ulaw_8000, alaw_8000, and pcm_8000.

**Who is it for?** ElevenLabs suits a team that wants to direct delivery with audio tags and will size a monthly plan to a campaign.

<div class="cmp-proscons" data-rival="elevenlabs"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>v4 Turbo and Flash at $0.04 per 1,000 characters</li><li>Audio tags such as [whispering] and [laughing] on Eleven 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>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 $0.04 for v4 Turbo and Flash, inside plans from Free to Business at $990 a month, and the Free plan is for non-commercial use.

## Which text to speech sounds most human on an outbound call?

Bland Speech v3 ranked 1st of 13 AI voices on Design Arena's Audio Realism Bench on 4 August 2026, at 1365, ahead of xAI's Grok TTS at 1156, Smallest AI's Lightning v3.1 Pro at 952, and ElevenLabs' Eleven v3 at 938.

Real human recordings scored 1500 in the same test, which [Design Arena published](https://notes.designarena.ai/audio-realism-benchmark-measuring-realism-in-audio-models/) with its method: vetted native speakers picking which of two clips sounds more like a real person. Deepgram's and Rime's voices weren't in those results, and models released since, such as ElevenLabs' Eleven v4, haven't been scored in them.

## How much does text to speech cost for 50,000 outbound calls?

Text to speech for 50,000 one-minute outbound calls costs between $600 and $2,500 in a month, from Grok TTS at $15 per million characters to Rime's Coda v2 at $0.05 a minute. Bland Speech costs $1,600 at $40 per million characters, in the middle of that range.

A cheaper rate is a real saving at this scale. Grok TTS saves $1,000 over Bland Speech on 40 million characters and Smallest AI $820, and the trade is a voice that ranked 3rd rather than 1st in Grok TTS's case and 11th at 952 in Smallest AI's. Enterprise pricing is custom, set with Bland's sales team.

Bland's [comparison of text to speech APIs](/compare/text-to-speech) lists more providers by job, and the [best-lists family](/compare/text-to-speech/best-lists) has the other ranked pages. A team weighing the voice of its sales reps can read [Best voice cloning for sales calls](/blog/best-voice-cloning-for-sales-calls).

## Frequently asked questions

### What is the best text to speech for outbound calls?

Bland Speech is the best text to speech for outbound calls. Bland Speech v3 ranked first of 13 AI voices on Audio Realism Bench on 4 August 2026, its WebSocket takes language model text as it is written, and it costs $0.04 per 1,000 characters with no monthly plan.

### How much does text to speech cost per outbound call?

A one-minute call with 800 characters of speech costs 3.2 cents on Bland Speech, 1.2 cents on Grok TTS, about 1.6 cents on Smallest AI's Lightning v3.1 Pro, and 5 cents on Rime's Coda v2. Each is that provider's published rate applied to one minute of speech.

### Does Bland Speech charge for the weeks a campaign is quiet?

No. Bland Speech bills $0.04 per 1,000 characters as they are used, with no monthly plan, and its credits do not expire, so a month with no calls costs nothing.

### What is the cheapest text to speech for outbound calling?

Grok TTS at $15 per million characters has the lowest rate here, followed by Smallest AI's Lightning v3.1 Pro at $19.50, against $40 for Bland Speech. Bland Speech is the pick when the first sentence a stranger hears should sound like a person, since it ranked first of 13 AI voices on Audio Realism Bench on 4 August 2026.

### Can an outbound agent use a sales rep's voice?

Yes, with the rep's consent. Bland Speech makes an instant clone from one clean sample of about 10 seconds, keeps it private to the customer's organization, and has the customer confirm the voice is theirs to use. Grok TTS clones from a reference clip in its console, Smallest AI from 5 to 15 seconds, and Rime's cloning is an Enterprise order that needs 30 to 60 minutes of recordings.
