Does speechEngine SDK Embedded run without an internet connection?

Yes. speechEngine SDK Embedded synthesizes speech entirely on the device, with no network connection required. All processing and data stay on the hardware, so speech output is immediate and consistent even in offline or constrained environments.

Which operating systems does it support?

It supports iOS, Android, and Embedded Linux out of the box. Other embedded operating systems can be supported on request. Reach out to our team with your target platform.

What hardware does it need to run?

It runs on ARM 32/64-bit processors. Neural Standard voices need a 64-bit ARM CPU at 1 GHz or higher; Neural Lite voices run from 800 MHz. The SDK uses 6 to 30 MB of RAM and 20 to 30 MB at runtime.

How much storage does each voice use?

Each voice uses 6 to 32 MB for Neural Lite and 9 to 35 MB for Neural Standard. Both tiers use Deep Neural Network technology, so you get natural sound at an embedded footprint.

Which programming languages can I integrate with?

You can integrate in C/C++ on Embedded Linux, Objective-C on iOS, and Java on Android. Each SDK package includes the APIs, sample code, and documentation to get started.

Can I control pronunciation and prosody?

Yes. Use SSML to adjust prosody, insert pauses, add phonetic transcriptions, and switch voices or languages mid-text. A per-language user dictionary with IPA support lets you define custom pronunciations for domain-specific terms.

What audio formats does it output?

It outputs 8 audio formats, including 16-bit linear PCM, 8-bit A-law and μ-law, 4-bit Dialogic ADPCM, and several Wave variants. This lets you match the audio pipeline of your target device.

How is it different from speechEngine SDK for desktop?

speechEngine SDK Embedded is built for on-device, small-footprint systems on ARM and mobile hardware. speechEngine SDK targets desktop and PC-based applications on Windows and Linux. See speechEngine SDK if you're building for desktop.

Listen

ReadSpeaker speechEngine SDK Embedded

On-device text-to-speech for embedded and mobile devices

ReadSpeaker speechEngine SDK Embedded adds natural AI voices to ARM-based devices, mobile apps, and embedded systems. It runs fully on-device, with no network connection required. 150+ AI voices across 50+ languages for iOS, Android, and Embedded Linux.

Talk to our team

Runs fully on-device
ARM 32/64-bit
150+ voices · 50+ languages

main.c · illustrative


      1 /* On-device TTS, no network */

      2 #include "rs_embedded_tts.h"

      3 

      4 int main() {

      5   rsEngine *e = rsEngineCreate("license.dat");

      6   rsEngineLoadVoice(e, "en-US-neural-standard");

      7   rsEngineSpeak(e, text, RS_AUDIO_PCM16);

      8   rsEngineDestroy(e);

      9 }

Why on-device

Speech that works with or without a connection

Many devices can't rely on the cloud. A medical monitor, a car, or an industrial controller has to speak the moment it's needed, every time. speechEngine SDK Embedded synthesizes speech entirely on the device, so audio output stays consistent and immediate even with no network. Your data never leaves the hardware, and latency never depends on a signal.

Built for embedded

Natural voice on hardware most TTS can't fit

A small footprint, by design

Each voice uses 6 to 35 MB of storage and runs in 20 to 30 MB of memory. That fits the constrained hardware embedded products are built on, from ARM 32-bit controllers to 64-bit mobile chips, without trading away natural sound.

150+ AI voices in 50+ languages

speechEngine SDK Embedded draws on ReadSpeaker's catalog of more than 150 AI voices across over 50 languages. New voices and languages are added regularly, so the same SDK can ship a product in one market or fifty.

Tune every word it reads

Adjust speaking rate, pitch, and volume to match the device. Use SSML (Speech Synthesis Markup Language) to insert pauses, switch voices or languages mid-sentence, and add phonetic transcriptions. A per-language user dictionary with IPA support lets you fix the pronunciation of product names and domain terms once, everywhere.

Technical specs

Built for constrained hardware

speechEngine SDK Embedded is engineered for small-footprint systems. The figures below let you size your hardware before you commit.

speechEngine SDK Embedded technical specifications
Specification	Value
Operating systems	iOS · Android · Embedded Linux (other embedded OS on request)
CPU	ARM 32/64-bit. Neural Standard: 1 GHz or higher (64-bit only) · Neural Lite: 800 MHz or higher
RAM	6 to 30 MB
Runtime memory	20 to 30 MB (Neural Lite and Neural Standard)
Voice footprint	6 to 32 MB (Neural Lite) · 9 to 35 MB (Neural Standard), Deep Neural Networks technology
Audio output	8 formats: 16-bit linear PCM, 8-bit A-law, 8-bit μ-law, 8-bit unsigned linear PCM Wave, 4-bit Dialogic ADPCM, and Wave variants
Text input	Plain text (multibyte) · SSML (UTF-8)
Documentation	Product Specification (PDF)

Neural Standard & Neural Lite

Two voice tiers, one SDK

Pick the balance that fits your hardware. Neural Standard runs on 64-bit ARM at 1 GHz and up, with voices from 9 to 35 MB for the richest sound. Neural Lite runs on ARM from 800 MHz, with voices from 6 to 32 MB for the tightest devices. Both use Deep Neural Network technology.

Neural Lite and Neural Standard compared
	Neural Lite	Neural Standard
CPU	ARM 800 MHz+	ARM 1 GHz+ (64-bit only)
Voice footprint	6 to 32 MB	9 to 35 MB
Technology	Deep Neural Networks	Deep Neural Networks

Platforms

Integrate in the language you already use

speechEngine SDK Embedded ships with the APIs, sample code, and documentation to integrate quickly. Build in C/C++ on Embedded Linux, Objective-C on iOS, or Java on Android. Each delivery includes the speech engine libraries, voice files, and integration docs for your licensed platform, plus support from our team.

Supported platforms and their integration languages
Platform	Language
Embedded Linux	C / C++
iOS	Objective-C
Android	Java

Integration

A few calls to first speech

Initialize the engine, load a voice, and synthesize. The SDK handles the rest on the device.

/* Illustrative - confirm exact API against the SDK docs before use */
#include "rs_embedded_tts.h"

int main() {
    rsEngine *engine = rsEngineCreate("license.dat");
    rsEngineLoadVoice(engine, "en-US-neural-standard");
    rsEngineSetRate(engine, 1.0);
    rsEngineSpeak(engine, "Welcome aboard.", RS_AUDIO_PCM16);
    rsEngineDestroy(engine);
    return 0;
}

Illustrative shape only. Refer to the SDK documentation and sample code for the exact API signatures of your delivered platform.

Where it runs

Voices for mission-critical devices

From mobile apps to mission-critical hardware, speechEngine SDK Embedded gives products a reliable voice wherever they run, including defense and other constrained environments.

Medical & industrial

Spoken alerts and readouts on monitors and controllers that can't wait for the cloud.

See healthcare solutions

Transportation

Navigation, turn-by-turn directions, and in-vehicle prompts that stay responsive even when the vehicle loses signal.

See transportation solutions

Accessibility (AAC & DAISY)

Natural speech for AAC communication devices and DAISY digital talking books.

See assistive technology solutions

Robotics

On-device speech that gives interactive toys and service robots a natural voice for more engaging interaction.

See embedded & on-device solutions

Announcement systems

Clear voice output for alarm and emergency notification systems.

Mobile devices

Audiobooks, language learning, and mobile apps that speak offline, with no network connection required.

Yes. speechEngine SDK Embedded synthesizes speech entirely on the device, with no network connection required. All processing and data stay on the hardware, so speech output is immediate and consistent even in offline or constrained environments.
It supports iOS, Android, and Embedded Linux out of the box. Other embedded operating systems can be supported on request. Reach out to our team with your target platform.
It runs on ARM 32/64-bit processors. Neural Standard voices need a 64-bit ARM CPU at 1 GHz or higher; Neural Lite voices run from 800 MHz. The SDK uses 6 to 30 MB of RAM and 20 to 30 MB at runtime.
Each voice uses 6 to 32 MB for Neural Lite and 9 to 35 MB for Neural Standard. Both tiers use Deep Neural Network technology, so you get natural sound at an embedded footprint.
speechEngine SDK Embedded supports more than 150 AI voices across over 50 languages. New voices and languages are added regularly. A complete list is available on the ReadSpeaker website.
You can integrate in C/C++ on Embedded Linux, Objective-C on iOS, and Java on Android. Each SDK package includes the APIs, sample code, and documentation to get started.
Yes. Use SSML to adjust prosody, insert pauses, add phonetic transcriptions, and switch voices or languages mid-text. A per-language user dictionary with IPA support lets you define custom pronunciations for domain-specific terms.
It outputs 8 audio formats, including 16-bit linear PCM, 8-bit A-law and μ-law, 4-bit Dialogic ADPCM, and several Wave variants. This lets you match the audio pipeline of your target device.
speechEngine SDK Embedded is built for on-device, small-footprint systems on ARM and mobile hardware. speechEngine SDK targets desktop and PC-based applications on Windows and Linux. See speechEngine SDK if you're building for desktop.