What is speechCloud API?

speechCloud API (SCAPI) is a cloud-hosted text-to-speech REST API from ReadSpeaker. You send text to the ReadSpeaker servers via HTTPS and receive back high-quality audio in your choice of format (MP3, Ogg, PCM, WAV, A-law or μ-law). It supports 90 languages and 300 voices, SSML markup, custom pronunciation dictionaries and timing data for synchronized highlighting. It is server-based with no software to install; integration is done via API key and HTTP requests.

How is speechCloud API different from webReader or docReader?

webReader and docReader are end-user products with their own user interface. webReader puts a Listen button on the HTML pages of your website; docReader does the same for the documents you publish. speechCloud API has no user interface. You call it server-to-server from your application to generate audio. Use it for connected devices, telephony and IVR, mobile apps, e-learning platforms that produce audio, audiobooks, or any product where you need text-to-speech inside your own software.

Which audio formats does the API return?

speechCloud API returns audio in six formats: MP3 (configurable bitrate 16, 24, 32, 48, 64 or 128 kbit/s), Ogg VBR, μ-law, A-law, raw PCM and WAV. PCM and WAV support sample rates of 22050, 16000 or 8000 Hz at 16-bit depth. The choice of format and audio parameters is set per request via the API parameters.

What is the maximum text size per request?

Each request takes up to 5,000 characters by default, counting everything except whitespace. The standard service is sized for up to 25,000 requests a day, with hourly peaks up to 2,000. Running higher or sustained traffic? Talk to ReadSpeaker about a dedicated configuration built for your volume.

Does speechCloud API support SSML and custom pronunciation?

Yes. SSML input is supported, including pauses, phonetic transcriptions, and inline switching of voice or language within a single text. Custom pronunciation dictionaries are also available: each customer can edit a voice-specific dictionary in the customer portal to control how brand names, technical terms or unusual words are pronounced.

Can I get timing data for synchronized highlighting?

Yes. speechCloud API can return RSDS which combines the audio and the timing information in a single response, saving credits compared to two separate requests. RSDS is currently available with MP3 at 48 kbit/s and is not available for every language or voice; contact ReadSpeaker for the current list.

How is speechCloud API priced?

speechCloud API runs on credits: one credit covers one character, counting everything except whitespace, plus any characters added by your pronunciation dictionary. You order production credits through your ReadSpeaker Account Manager, and new accounts usually start with trial credits. A per-second-of-speech model is available on request. Capacity can be licensed by characters, requests, speech time, concurrent channels, or generation speed. Contact ReadSpeaker for a quote matched to your volume.

Is speechCloud API secure and compliant?

ReadSpeaker as an organization is ISO/IEC 27001:2022 certified, and GDPR compliant. All API traffic uses HTTPS. For EU customers and cross-border data transfers, ReadSpeaker relies on Standard Contractual Clauses (SCCs). Per the ReadSpeaker Privacy Notice, ReadSpeaker's synthesised voices are blended voice models that do not constitute Personal Information under GDPR. Audio data must not be stored, reused or redistributed unless explicitly agreed in the order confirmation.

What is the latency? Can I use speechCloud API for real-time use cases?

speechCloud API supports real-time use cases. By default, requests use streaming mode (streaming=1), where the response redirects to a streaming server that begins delivering audio as soon as possible, for the shortest time-to-audio. For batch delivery, set streaming=0 and the file is returned once fully generated. The number of concurrent real-time channels is part of your licensed capacity and can be scaled in the order confirmation.

How do I get started? Where do I get an API key?

Create an account with ReadSpeaker. New accounts typically receive a limited amount of trial credits so you can evaluate speechCloud API. Your API key is available in the SCAPI customer portal and must be included with each request. Implementation and user instructions, sample code in popular programming languages, and your editable pronunciation dictionary are also accessible from the portal. Contact ReadSpeaker to start your account.

Listen

ReadSpeaker speechCloud API

Text-to-speech API for any app, device or workflow

Send text to one REST endpoint, get back natural speech in 90 languages and 300 voices. SSML, custom pronunciation, and six audio formats, ready for production from the first call.

Start a trial Talk to sales

Already a customer? Sign in to the Customer Portal

90 languages
300 voices
EU based, GDPR, ISO 27001

POST · speechCloud API

# curl example — request synthesised speech
curl -X POST "https://tts.readspeaker.com/a/eg/v2.0" \
  -d "key=YOUR_API_KEY" \
  -d "voice=Adam" \
  -d "lang=en_uk" \
  -d "audioformat=mp3" \
  -d "audiobitrate=48" \
  -d "text=Welcome to ReadSpeaker." \
  --output speech.mp3

200 OKaudio/mpeg · 48 kbps

Trusted by 12,000 customers worldwide

Try a voice

Listen to what your API will return

Four voices across four languages. Same engine you'll integrate via the API.

Adam

Male

English (United Kingdom)

Roxane

Female

Français (France)

Lena

Female

Deutsch (Deutschland)

Sayaka

Female

日本語 (Japan)

Want a custom voice or a language not listed? Browse all 300 voices →

Customer stories

Brands building voice products with ReadSpeaker

Sonos

ReadSpeaker's approach to custom voice is innovative and extremely thorough – exactly what we required for this project.

Read the full story

Cloud-based speech generation

Real-time or batch, anywhere your product needs a voice

speechCloud API is a cloud-based, server-to-server text-to-speech service. Call it inside your own application to generate audio on the fly for end-users, or to produce audio in bulk for later playback. It replaces costly manual recordings and runs around the clock, so you never host or scale a text-to-speech engine yourself.

Real-time apps & devices

Add text-to-speech to a mobile app, a connected car, a medical or industrial appliance, a wearable or a kiosk. Stream audio on demand without bundling a TTS engine on-device, so footprint and battery stay low. Receive natural speech in 90 languages with a single REST call.

Mobile
IoT
Kiosks
Wearables
Connected cars
Medical devices

See embedded & on-device solutions

Batch audio production

Generate audio in bulk for e-learning courses, assessments, training content, audiobooks, audio guides and dialog for movies or animations. Streamed audio can be served back to end-users from your own infrastructure when allowed by your contract. Custom pronunciation dictionaries make sure brand names and technical terms read correctly across every track.

e-learning
Audiobooks
Assessments
Training
Audio guides
Movies & animation

See education solutions

Telephony, IVR & voice agents

Power phone systems, contact centers and LLM-powered voice agents with natural voices. A-law and μ-law output formats are the G.711 codecs used in telephony systems. Use SSML to control pacing, switch language inline and pronounce brand names via your custom dictionary. Concurrent real-time channels scale with your contract.

IVR
G.711
Voice agents
SSML
A-law / μ-law

See conversational AI solutions

Features

Production text-to-speech, end to end

Everything your product needs to talk back.

90 languages, 300 voices

Pick from a deep, professional voice catalog per language. Voices are licensed individually; bring more on at any time.

SSML input

Use SSML for pauses, phonetic transcriptions, and inline switching of voice or language within a single text. Standard SSML markup is supported.

Custom pronunciation dictionaries

Each customer can edit a voice-specific dictionary in the customer portal, to control how brand names, technical terms and unusual words are pronounced. ReadSpeaker linguists can correct mispronunciations on request.

Timing data (RSDS)

RSDS returns audio and timing markers in a single response, saving credits compared to two separate calls. Useful for word and sentence highlighting in your UI. Available in MP3 48 kbps only, not on all voices.

Six audio output formats

MP3 (16, 24, 32, 48, 64, 128 kbit/s), Ogg VBR, μ-law, A-law, raw PCM, WAV. PCM and WAV at 22050, 16000 or 8000 Hz, 16-bit depth. Set per request via API parameters.

Streaming or download

Set streaming=1 to start playback the moment audio is ready, ideal for live use. Or take the finished file as a single download.

REST API over HTTPS

API key included with each call. Query available languages, voices, current usage and remaining credits via the same REST interface.

SCAPI customer portal

Edit your pronunciation dictionary, track usage and remaining credits, and roll a new API key the moment misuse is suspected, all from one portal.

Hosted & redundant infrastructure

Hosted by ReadSpeaker in a high-security environment with redundancy and failover. Nothing to download or install.

Try it in your stack

From text to audio in one request

A look at what a call looks like in four languages. Authenticate with your API key, sent with every request.

# Synthesise a sentence to MP3 (48 kbps)
curl -X POST "https://tts.readspeaker.com/a/eg/v2.0" \
  -d "key=YOUR_API_KEY" \
  -d "voice=Adam" \
  -d "lang=en_uk" \
  -d "audioformat=mp3" \
  -d "audiobitrate=48" \
  -d "text=Welcome to ReadSpeaker." \
  --output speech.mp3

# With SSML, switching language inline
curl -X POST "https://tts.readspeaker.com/a/eg/v2.0" \
  -d "key=YOUR_API_KEY" \
  -d "voice=Adam" \
  -d "lang=en_uk" \
  -d "input=ssml" \
  --data-urlencode 'text=<speak>Hello, <voice name="Roxane" xml:lang="fr-FR">bonjour</voice>.</speak>' \
  --output greeting.mp3

# Python 3 — requests library
import requests

resp = requests.post(
    "https://tts.readspeaker.com/a/eg/v2.0",
    data={
        "key": "YOUR_API_KEY",
        "voice": "Adam",
        "lang": "en_uk",
        "audioformat": "mp3",
        "audiobitrate": 48,
        "text": "Welcome to ReadSpeaker.",
    },
)

if resp.status_code == 200:
    with open("speech.mp3", "wb") as f:
        f.write(resp.content)
else:
    print("Error:", resp.status_code, resp.text)

// Node.js 18+ — native fetch + fs/promises
import { writeFile } from "node:fs/promises";

const body = new URLSearchParams({
  key: "YOUR_API_KEY",
  voice: "Adam",
  lang: "en_uk",
  audioformat: "mp3",
  audiobitrate: "48",
  text: "Welcome to ReadSpeaker.",
});

const resp = await fetch("https://tts.readspeaker.com/a/eg/v2.0", {
  method: "POST",
  body,
});

if (!resp.ok) throw new Error(`SCAPI ${resp.status}`);
await writeFile("speech.mp3", Buffer.from(await resp.arrayBuffer()));

<?php
// PHP 8 — cURL
$ch = curl_init("https://tts.readspeaker.com/a/eg/v2.0");

curl_setopt_array($ch, [
    CURLOPT_POST          => true,
    CURLOPT_POSTFIELDS    => http_build_query([
        "key"          => "YOUR_API_KEY",
        "voice"        => "Adam",
        "lang"         => "en_uk",
        "audioformat"  => "mp3",
        "audiobitrate" => 48,
        "text"         => "Welcome to ReadSpeaker.",
    ]),
    CURLOPT_RETURNTRANSFER => true,
]);

$audio = curl_exec($ch);
$code  = curl_getinfo($ch, CURLINFO_HTTP_CODE);
curl_close($ch);

if ($code === 200) {
    file_put_contents("speech.mp3", $audio);
} else {
    error_log("SCAPI error: $code");
}

Endpoint shape is illustrative. Refer to the latest Product Specification and the Customer Portal documentation for the current endpoint URL and parameter names.

Technical specifications

Every spec your security review will ask about

Authentication, transport, language coverage, throughput limits and audio formats. The answers your security and ops teams need before integration.

speechCloud API technical specifications
Specification	Value
Authentication	API key, included with each request
Transport	REST over HTTPS
Languages supported	90 languages
Voices available	300 voices
Max text per request	5,000 characters (excluding whitespace), default
Intended usage limit	≤ 25,000 requests/day · peaks ≤ 2,000 requests within a single hour
Audio output formats	MP3 · Ogg VBR · μ-law · A-law · PCM · WAV
MP3 bitrates	16, 24, 32, 48, 64, 128 kbit/s
PCM / WAV sample rates	22050, 16000, 8000 Hz @ 16-bit
Input encoding	UTF-8 recommended; most common encodings supported
SSML support	Yes, pauses, phonetic transcriptions, inline voice/language switching
Timing data (RSDS)	Audio + timing in one response · currently MP3 48 kbit/s only · not all voices
Streaming mode	streaming=1 (default) → redirect to streaming server; streaming=0 → file delivered after full generation
Audio storage rules	No storage, reuse or redistribution unless explicitly agreed in order confirmation
Telephony integration	A-law and μ-law output formats (G.711 codecs)
Documentation	Product Specification (PDF)

Pricing

Pay for what you generate, scale when you need

speechCloud API uses a credits model. Production credits are ordered through your ReadSpeaker Account Manager and added to your account by the ReadSpeaker support team. New accounts typically receive a limited amount of trial credits for initial testing.

Request a quote

Start free: new accounts typically receive trial credits. Request access via the contact form.

Credits

speechCloud API

Credits model1 credit = 1 character, excluding whitespace

How credits work

1 credit = 1 character, excluding whitespace, including any changes from pronunciation dictionary entries
Per-second-of-speech variant available via order confirmation for predictable cost on long audio
Trial credits at account start so you can evaluate before committing

Licensed capacity is measured by

Number of characters over the contract period
Number of requests and total speech time
Concurrent real-time channels for telephony and live use cases
Generation speed when responsiveness is critical

Compliance

Compliance you can hand straight to your auditor

ReadSpeaker is an ISO/IEC 27001:2022 certified organization, GDPR compliant, with Standard Contractual Clauses for cross-border data transfers.

ISO/IEC 27001:2022 (ISO/IEC 27001:2022 — Information security management)
ReadSpeaker organization
GDPR (General Data Protection Regulation)
Privacy Notice published
EU SCCs (EU Standard Contractual Clauses)
For cross-border transfers
EU based (European Union based organization)
Headquartered in Sweden (ReadSpeaker AB, Uppsala)

Per the ReadSpeaker Privacy Notice, ReadSpeaker's synthesised voices are blended voice models that do not constitute Personal Information under GDPR.

speechCloud API (SCAPI) is a cloud-hosted text-to-speech REST API from ReadSpeaker. You send text to the ReadSpeaker servers via HTTPS and receive back high-quality audio in your choice of format (MP3, Ogg, PCM, WAV, A-law or μ-law). It supports 90 languages and 300 voices, SSML markup, custom pronunciation dictionaries and timing data for synchronized highlighting. It is server-based with no software to install; integration is done via API key and HTTP requests.
webReader and docReader are end-user products with their own user interface. webReader puts a Listen button on the HTML pages of your website; docReader does the same for the documents you publish. speechCloud API has no user interface. You call it server-to-server from your application to generate audio. Use it for connected devices, telephony and IVR, mobile apps, e-learning platforms that produce audio, audiobooks, or any product where you need text-to-speech inside your own software.
speechCloud API returns audio in six formats: MP3 (configurable bitrate 16, 24, 32, 48, 64 or 128 kbit/s), Ogg VBR, μ-law, A-law, raw PCM and WAV. PCM and WAV support sample rates of 22050, 16000 or 8000 Hz at 16-bit depth. The choice of format and audio parameters is set per request via the API parameters.
Each request takes up to 5,000 characters by default, counting everything except whitespace. The standard service is sized for up to 25,000 requests a day, with hourly peaks up to 2,000. Running higher or sustained traffic? Talk to ReadSpeaker about a dedicated configuration built for your volume.
Yes. SSML input is supported, including pauses, phonetic transcriptions, and inline switching of voice or language within a single text. Custom pronunciation dictionaries are also available: each customer can edit a voice-specific dictionary in the customer portal to control how brand names, technical terms or unusual words are pronounced.
Yes. speechCloud API can return RSDS which combines the audio and the timing information in a single response, saving credits compared to two separate requests. RSDS is currently available with MP3 at 48 kbit/s and is not available for every language or voice; contact ReadSpeaker for the current list.
speechCloud API runs on credits: one credit covers one character, counting everything except whitespace, plus any characters added by your pronunciation dictionary. You order production credits through your ReadSpeaker Account Manager, and new accounts usually start with trial credits. A per-second-of-speech model is available on request. Capacity can be licensed by characters, requests, speech time, concurrent channels, or generation speed. Contact ReadSpeaker for a quote matched to your volume.
ReadSpeaker as an organization is ISO/IEC 27001:2022 certified, and GDPR compliant. All API traffic uses HTTPS. For EU customers and cross-border data transfers, ReadSpeaker relies on Standard Contractual Clauses (SCCs). Per the ReadSpeaker Privacy Notice, ReadSpeaker's synthesised voices are blended voice models that do not constitute Personal Information under GDPR. Audio data must not be stored, reused or redistributed unless explicitly agreed in the order confirmation.
speechCloud API supports real-time use cases. By default, requests use streaming mode (streaming=1), where the response redirects to a streaming server that begins delivering audio as soon as possible, for the shortest time-to-audio. For batch delivery, set streaming=0 and the file is returned once fully generated. The number of concurrent real-time channels is part of your licensed capacity and can be scaled in the order confirmation.
Create an account with ReadSpeaker. New accounts typically receive a limited amount of trial credits so you can evaluate speechCloud API. Your API key is available in the SCAPI customer portal and must be included with each request. Implementation and user instructions, sample code in popular programming languages, and your editable pronunciation dictionary are also accessible from the portal. Contact ReadSpeaker to start your account.

VoiceLab

Need a brand voice of your own?

speechCloud API ships with our professional voice catalog.

A brand voice built with you in studio

For a fully custom brand voice built with you in studio, explore VoiceLab. ReadSpeaker's computational linguists have been crafting brand voices since 1999.

Discover VoiceLab

Text-to-speech API for any app, device or workflow

Listen to what your API will return

Adam

Roxane

Lena

Sayaka

Brands building voice products with ReadSpeaker

Real-time or batch, anywhere your product needs a voice

Real-time apps & devices

Batch audio production

Telephony, IVR & voice agents

Production text-to-speech, end to end

90 languages, 300 voices

SSML input

Custom pronunciation dictionaries

Timing data (RSDS)

Six audio output formats

Streaming or download

REST API over HTTPS

SCAPI customer portal

Hosted & redundant infrastructure

From text to audio in one request

Every spec your security review will ask about

Pay for what you generate, scale when you need

speechCloud API

Compliance you can hand straight to your auditor

Product Specification

ReadSpeaker Privacy Notice

Questions about speechCloud API

Need a brand voice of your own?

A brand voice built with you in studio

Ready to put a voice in your product?