# 7 best text to speech APIs for logistics and delivery updates in 2026

By Samuel Oram-Jones

A delivery update is mostly digit strings, an order number, a door code, and a window of time, and a voice that reads a reference as one long number leaves the recipient asking for it again.

Seven text to speech APIs handle it with different tools, from a letter-by-letter spelling function to a switch that expands numbers into words. Bland Speech v3 ranked first of 13 AI voices in Design Arena's Audio Realism Bench results of 4 August 2026, and Amazon Polly Standard costs a tenth of its price per character, as of November 2026.

## What are the best text to speech APIs for logistics and delivery updates?

The best text to speech API for logistics and delivery updates in 2026 is Bland Speech, followed by Cartesia, Deepgram, Rime, Grok TTS, Gradium, and Amazon Polly, with Bland Speech first for updates a recipient can answer and Polly first for one-way alerts that cost the least per character.

- **Bland Speech:** Best overall for delivery updates a recipient can answer back
- **Cartesia:** Best for pronunciation dictionaries and a first-audio claim under 90ms
- **Deepgram:** Best for built-in text normalization in English plus 10 more languages
- **Rime:** Best for spelling out codes letter by letter with spell()
- **Grok TTS:** Best for a $15 list price with a normalization switch
- **Gradium:** Best for rewriting rules for numbers and custom patterns
- **Amazon Polly:** Best for a $4 list price on one-way alerts

| API | Number and ID handling | Price as the company publishes it | Audio Realism Bench, 4 August 2026 |
|---|---|---|---|
| Bland Speech | Spelled-out numbers, short digit runs, and pause markers | $0.04 per 1,000 characters | #1 AI voice, 1365 |
| Cartesia | Pronunciation dictionaries, and a claim that order numbers read exactly | 1 credit per character on monthly plans | Sonic 3.5: #6, 1096 |
| Deepgram | Built-in text normalization, and inline IPA in Early Access | $0.045 per 1,000 characters for Flux TTS | Wasn't in those results |
| Rime | spell() for letter-by-letter reading, and custom lexicons | $0.030 to $0.050 a minute | Wasn't in those results |
| Grok TTS | A text normalization switch for numbers, abbreviations, and symbols | $15 per million characters | #3, 1156 |
| Gradium | Rewriting rules for numbers, emails, and custom patterns | $36 to $58 per million credits | Wasn't in those results |
| Amazon Polly | SSML with phonetic pronunciation and pauses | $4 to $100 per million characters, by engine | Wasn't in those results |

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

- **Number and ID handling:** the tools each company documents for reading digits, codes, and unusual words aloud.
- **Price:** each company's list price with the model or plan named. Rime bills by the minute and says a minute is about 1,000 characters, and a Gradium credit is one character.
- **Audio Realism Bench:** the rank among 13 AI voices and the score in Design Arena's published results.

</details>

For updates a recipient answers, the voice that read the time window also has to stop when the recipient cuts in, and Bland Speech takes first place on its 4 August 2026 realism result and a WebSocket that handles that interruption, while Polly Standard sits at $4 per million characters for alerts nobody replies to.

## How does a text to speech API read tracking numbers and time windows?

A text to speech API reads a tracking number well when the writer can control how digits are grouped, spelled, and paused, and each of these seven offers a different way to do it.

### What does each API give for codes and pronunciation?

Bland advises digit runs of about six characters, a pause marker such as <|0.5|> to break a longer run, and a period at the end of the sentence. Spelling a number as words gives exact control, Bland's studio reads phone numbers in short groups with a pause between them, and Bland Speech reads an email address as "john at gmail dot com" without rewriting.

Cartesia accepts pronunciation dictionaries passed by id and says Sonic reads order numbers exactly. Deepgram's Flux TTS has built-in text normalization in English plus 10 languages, and Rime offers a spell() function and custom lexicons. Grok TTS has a text normalization switch that is off by default, Gradium applies rewriting rules for numbers and custom patterns, and Amazon Polly takes SSML with phonetic pronunciation and pauses.

### What does a month of delivery updates cost?

Take 100,000 updates a month at about 400 characters each, which is 40 million characters. At each company's list price, Amazon Polly Standard costs $140 after its free 5 million characters, Grok TTS $600, Rime's Mist v3 about $1,200, Gradium about $1,500 on its L plan, Cartesia about $1,515 on Scale with overage, Bland Speech $1,600, Deepgram's Flux TTS $1,800, and Rime's Coda v2 about $2,000.

Polly's price is the lowest by a wide margin, and for a one-way alert that only needs a clear reference it is a fair choice. The extra cost of Bland Speech buys a voice that scored 1365 in the blind tests Design Arena published on 4 August 2026, and a WebSocket where a new context stops the turn that is playing.

## How do the 7 best text to speech APIs for logistics and delivery updates compare?

These seven differ most in how they read numbers and what they charge at volume.

### 1. Bland Speech: Best overall for delivery updates a recipient can answer back

**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. It streams over HTTP or over a WebSocket that takes text as a language model writes it, where a new context preempts the turn that is playing, and it returns PCM from 8 to 48 kHz or mu-law at 8 kHz.

**Who is it for?** Bland Speech suits logistics teams whose delivery calls go both ways, where a recipient hears a window and answers with a reschedule or an access code.

<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>Pause markers from 0.1 to 10.0 seconds for breaking up long references</li><li>A new context on the WebSocket preempts the turn that is playing</li><li>Mu-law at 8 kHz for phone lines</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>More than a one-way alert script 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, and the [text to speech API](/text-to-speech-api) page has the endpoints.

### 2. Cartesia: Best for pronunciation dictionaries and 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 returns mu-law or A-law at 8 kHz, and a pronunciation dictionary of IPA or sounds-like entries is passed by id with each request.

**Who is it for?** Cartesia suits teams that want a reusable dictionary of street names and product codes and a first-audio claim under 90ms.

<div class="cmp-proscons" data-rival="cartesia"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Pronunciation dictionaries with IPA or sounds-like entries</li><li>Word timestamps on the WebSocket</li><li>Cartesia says first audio arrives in under 90ms</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>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. Sonic 3.6 came out after the 4 August 2026 results.

### 3. Deepgram: Best for built-in text normalization in English plus 10 more languages

**What does it do?** Deepgram sells speech recognition, text to speech, and a voice agent API. Flux TTS streams over a WebSocket, ships in English plus 10 languages with built-in text normalization, and returns mu-law or A-law at 8 kHz.

**Who is it for?** Deepgram suits teams that want normalization handled by the model and speech recognition from the same vendor.

<div class="cmp-proscons" data-rival="deepgram"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Built-in text normalization in English plus 10 languages</li><li>Inline IPA pronunciation in Early Access</li><li>Pay as you go with $200 of free credit</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Deepgram's 40+ additional languages need a team's own text normalization layer</li><li>Its voices weren't in Audio Realism Bench's results of 4 August 2026</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. [Bland Speech vs Deepgram](/compare/bland-speech-vs-deepgram) compares the two.

### 4. Rime: Best for spelling out codes letter by letter with spell()

**What does it do?** Rime sells conversational voice models for voice agents, IVR, and telephony. A spell() function reads a code letter by letter, Coda accepts a custom lexicon of X-SAMPA pronunciations, and audio returns as G.711 mu-law at 8 kHz.

**Who is it for?** Rime suits teams reading alphanumeric references aloud and paying by the minute.

<div class="cmp-proscons" data-rival="rime"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>spell() for letter-by-letter reading</li><li>Custom lexicons for English, German, Spanish, French, Italian, and Portuguese on Coda</li><li>Flex plan with no monthly fee</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Custom pauses with angle-bracket markers work on Mist v3 only</li><li>Its voices weren't in Audio Realism Bench's results of 4 August 2026</li></ul></div></div>

**Pricing:** $0.030 a minute for Mist v3 and $0.050 for Coda v2 on Flex, with $50 of free usage every month, and Scale at $100 a month takes 20% off every rate.

### 5. Grok TTS: Best for a $15 list price with a normalization switch

**What does it do?** Grok TTS is a text to speech API from xAI, a company now named SpaceXSI. A text normalization setting expands written numbers, abbreviations, and symbols into spoken form, a replace setting takes respellings or IPA, and audio returns as mu-law or A-law at 8 kHz.

**Who is it for?** Grok TTS suits teams that want normalization as a request setting at a low per-character price.

<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>Pronunciation replacements by respelling or IPA</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>Text normalization is off by default and turned on per request</li><li>Character-level timestamps are interpolated when normalization expands a symbol</li></ul></div></div>

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

### 6. Gradium: Best for rewriting rules for numbers and custom patterns

**What does it do?** Gradium sells real-time voice AI for voice agents. Its text rewriting rules normalize numbers, emails, and custom patterns, a 9 November 2026 release improved number pronunciation in Spanish and Portuguese, and audio returns as mu-law or A-law at 8 kHz.

**Who is it for?** Gradium suits teams that want to write their own rewrite rules for reference formats, in English, French, German, Spanish, or Portuguese.

<div class="cmp-proscons" data-rival="gradium"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Rewriting rules for numbers, emails, and custom patterns</li><li>Break tags for timed pauses</li><li>A WebSocket that takes text token by token</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Five languages: English, French, German, Spanish, and Portuguese</li><li>Unused plan credits don't roll over to the next month</li></ul></div></div>

**Pricing:** credits at 1 per character, with paid plans from about $12 a month and rates from $58 down to $36 per million credits.

### 7. Amazon Polly: Best for a $4 list price on one-way alerts

**What does it do?** Amazon Polly offers four voice engines, Generative, Long-Form, Neural, and Standard. SSML covers phonetic pronunciation and pauses, audio returns as mu-law or A-law at 8 kHz, and Amazon says generated speech can be cached and replayed at no additional cost.

**Who is it for?** Polly suits logistics teams on AWS that send large volumes of templated alerts and weigh cost per character above everything else.

<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, with 5 million free each month</li><li>Cached and replayed speech at no additional cost</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Its voices weren't in Audio Realism Bench's results of 4 August 2026</li><li>Each request takes up to 3,000 billed characters</li></ul></div></div>

**Pricing:** $4 per million characters for Standard voices, $16 for Neural, $30 for Generative, and $100 for Long-Form.

## Which text to speech sounds most like a person on a delivery 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. Grok TTS ranked third at 1156 and Cartesia's Sonic 3.5 sixth at 1096, while Deepgram, Rime, Gradium, and Amazon Polly weren't in those results.

Design Arena built its [blind tests](https://notes.designarena.ai/audio-realism-benchmark-measuring-realism-in-audio-models/) from private prompts of up to about 40 words, phone-agent lines among them, so the scores say how human a voice sounded and not how it read a tracking number.

## How were these text to speech APIs compared?

The seven were compared on how each reads numbers and codes, price at volume, phone audio, and voice realism. Each rival detail comes from that company's own pricing and documentation, voice scores come from Design Arena's results of 4 August 2026, and the monthly example assumes 400 characters an update.

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 8 kHz phone audio covers the audio format these calls use.

## Frequently asked questions

### What is the best text to speech API for logistics and delivery updates?

Bland Speech is the best text to speech API for logistics and delivery updates that a recipient can answer, in 2026. Bland Speech v3 ranked first of 13 AI voices on Audio Realism Bench on 4 August 2026, and its WebSocket stops the playing turn when a new context arrives. Amazon Polly Standard costs less for one-way alerts.

### How do these APIs handle order numbers and tracking references?

Each company documents its own tools. Bland Speech takes spelled-out numbers and pause markers, Cartesia takes pronunciation dictionaries, Deepgram's Flux TTS normalizes text in English plus 10 languages, Rime has spell(), Grok TTS has a normalization setting, Gradium has rewriting rules, and Polly takes SSML with phonetic pronunciation.

### Which text to speech API costs the least for bulk alerts?

Amazon Polly Standard voices cost $4 per million characters, with 5 million free each month, and Grok TTS costs $15 per million. Bland Speech costs $40 per million, and Deepgram's Flux TTS costs $45.

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

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

### How does Bland Speech pause inside a long reference?

A pause marker such as <|0.5|> inserts a pause from 0.1 to 10.0 seconds, and Bland says to budget roughly a second more than the marker asks for, because the marker splits the line into two segments.

### What does Bland Speech cost for 100,000 delivery updates?

At about 400 characters an update, 100,000 updates is 40 million characters, which costs $1,600 at $0.04 per 1,000 characters. A new account gets 50,000 characters free once.
