Best Text-to-Speech Engines for Android (and How to Switch)
Android's voice comes from a swappable engine. Google's is the default; RHVoice, SherpaTTS, Vocalizer and Acapela are the main alternatives. How to switch.
Key takeaways
- On Android, apps such as TalkBack, Select to Speak, @Voice and Moon+ Reader speak through a system text-to-speech engine that you choose in Settings.
- Google's free engine, now called Speech Recognition & Synthesis, is the best starting point for most people. Install the voice data for your language and try each voice before switching engines.
- For fully offline, open-source neural voices, SherpaTTS on F-Droid runs Piper and Coqui models on your phone. RHVoice and eSpeak NG are lighter open-source options.
- Paid engines such as Vocalizer (Code Factory) and Acapela sell voices one at a time, with free trials.
- Apps like Speechify, ElevenReader and Frateca use their own cloud voices, so changing the system engine doesn't affect them.
Android doesn’t have one built-in voice. It has a text-to-speech engine, a separate app that turns text into speech for any other app that asks. TalkBack, Select to Speak, many ebook readers and navigation prompts all send their text to whichever engine you’ve chosen. Change the engine, or the voice inside it, and all of those apps sound different.
For most people, the best engine is the one already on the phone: Google’s Speech Recognition & Synthesis, with the right voice data installed. If you want something different, this guide compares the main alternatives and shows how to switch, step by step. Details were checked against official help pages and app listings on November 12, 2026.
Quick recommendations
| You want… | Try |
|---|---|
| The best free all-rounder | Google Speech Recognition & Synthesis, with voice data installed |
| Open-source neural voices, fully offline | SherpaTTS (F-Droid), running Piper or Coqui voices |
| A light open-source engine for Russian, Ukrainian, Polish and similar languages | RHVoice |
| A very fast, robotic voice for screen-reader power users | eSpeak NG |
| Polished commercial voices in many languages | Vocalizer TTS (Code Factory) or Acapela TTS Voices |
| Natural cloud voices for long documents | A reader app with its own voices, such as Speechify, ElevenReader or Frateca |
How to change the text-to-speech engine
On a Pixel or stock Android 15 and 16
- Open Settings → Accessibility → Text-to-speech output.
- Tap Preferred engine and choose the engine you want.
- Tap the gear icon next to the engine, then Install voice data, and download your language.
- Back on Text-to-speech output, set Language, Speech rate and Pitch.
- Tap Play to hear a sample. Reset puts rate and pitch back to normal.
Google’s help page describes the same screen and notes that the engine list can include Google’s own, your phone maker’s, and any third-party engines you’ve installed from Google Play.
On a Samsung Galaxy
Samsung’s own help pages give different locations for this setting depending on the region and software version: Settings → General management → Text-to-speech on the US page, Settings → Accessibility → Text-to-speech on others. The fastest way is to open Settings, tap the search icon and type Text-to-speech. The screen has the same options: preferred engine, language, speech rate and pitch.
Samsung phones offer both Samsung’s engine and Google’s, with different voices and languages. Since One UI 7 and Android 15, app developers such as FBReader and many users report that Samsung’s voices only work in Samsung’s own apps. Samsung hasn’t announced this officially, but if a reading app goes silent or falls back to a different voice, select Google’s engine.
The engines compared
| Engine | Developer | Cost | Works offline | Open source | Get it from |
|---|---|---|---|---|---|
| Speech Recognition & Synthesis | Free | Yes, with voice data installed | No | Preinstalled / Google Play | |
| Samsung TTS | Samsung | Free | Yes | No | Preinstalled on Galaxy phones |
| SherpaTTS | woheller69 | Free | Yes, after a one-time model download | Yes (GPL-3.0) | F-Droid |
| RHVoice | RHVoice project | Free; a few voices are paid | Yes | Yes (GPL-3.0+) | Google Play, F-Droid |
| eSpeak NG | eSpeak NG project | Free | Yes | Yes (GPL-3.0+) | F-Droid |
| Vocalizer TTS | Code Factory | Per-voice purchase, 7-day trial | Yes | No | Google Play |
| Acapela TTS Voices | Acapela Group | About €3.99 per voice, free trial | Yes, once bought | No | Google Play |
| Eloquence TTS | Code Factory | Subscription after a 7-day trial | No, needs internet | No | Google Play |
Google Speech Recognition & Synthesis
Google’s engine is free, preinstalled on most Android phones, and installed more than 10 billion times on Google Play. It used to be called Speech Services by Google. Google’s TalkBack documentation lists 69 languages and locales for it, with the caveat that not every device supports every language in every region.
The difference between a robotic and a natural Google voice is usually the voice data. Open the engine’s settings, install your language, then try each voice; the variants within a language can sound quite different. Downloaded voices work offline.
SherpaTTS: neural voices that run on your phone
SherpaTTS is a free, open-source engine that runs modern neural voice models, from the Piper and Coqui projects, directly on the phone. It downloads the voice model once and then, in the developer’s words, works entirely offline. It’s available on F-Droid, needs Android 10 or later, and isn’t on Google Play.
The voices are much more natural than older offline engines. Larger models take more storage and processing, so test on your own phone. The developer has warned that Google’s new developer verification rules, which begin in some countries from September 2026 and roll out globally in 2027, may stop it working on certified Android devices, so keep an eye on its F-Droid page.
Advanced users can also install engine builds from the sherpa-onnx project on GitHub, which support Piper, Kokoro, Matcha and other voice models.
RHVoice
RHVoice is an open-source engine that’s particularly strong for Russian, Ukrainian, Polish, Georgian and other languages that commercial engines cover less well, along with English and Brazilian Portuguese. It’s free, with a few paid voices, and runs offline. Its voices sound synthetic but clear, and it’s light on resources.
eSpeak NG
eSpeak NG is the classic open-source formant synthesizer, covering more than 100 languages. It sounds robotic, but it stays crisp at very high speeds, which is why some experienced screen-reader users prefer it. The official build is on F-Droid; a Google Play version from another developer says it is not the official project and includes ads.
Vocalizer and Acapela
Vocalizer TTS from Code Factory offers over 120 voices in more than 50 languages, bought one voice at a time after a 7-day trial. The app had its last update in October 2024. Acapela TTS Voices sells individual voices for about €3.99 each, plus a premium children’s voice pack, with a free online trial.
Both run offline once you’ve bought a voice. They’re worth a look if you need a particular language or accent that Google’s engine handles poorly.
Engines that are fading
CereProc’s CerePlay app now says it’s deprecated and won’t be updated. Code Factory’s Eloquence, the classic voice some blind users prefer, is back on Google Play as a subscription and needs an internet connection.
Which apps use the engine you choose
Apps that speak through the system engine follow whatever you set in Text-to-speech output:
- TalkBack and Select to Speak, Android’s own accessibility tools.
- @Voice Aloud Reader, which uses the voices on your device, including Google’s and Samsung’s, and can also connect to cloud voices with your own API key.
- Moon+ Reader and Librera, popular ebook readers.
Some Google apps ignore it. Code Factory and Acapela both note that Google Maps and Gemini only use Google’s own engine.
Reader apps with their own voices don’t use the system engine at all. Speechify, ElevenReader and Frateca generate speech with their own cloud voices, so they sound the same whatever engine your phone uses. That’s the practical route if you want a natural voice for long documents rather than for the whole phone. The best text-to-speech app for Android compares them, and how to make Android read text aloud covers the built-in readers.
Fixing common problems
The voice sounds robotic. Install the voice data for your language in the engine’s settings and try each voice. Make sure the engine you set up is selected as the preferred engine.
Your language is missing. Not every engine supports every language. Install a second engine that does; Android lets you keep several and pick one as preferred.
A newly installed engine doesn’t appear. Open the engine app once, then return to Text-to-speech output. Restarting the phone usually helps.
A reading app went silent after an update. On Samsung phones, switch the preferred engine to Google’s. On any phone, check that the app has its own voice setting that might override the system choice.
Speech is too fast or slow everywhere. Adjust Speech rate in Text-to-speech output, then fine-tune inside individual apps if they have their own speed control.
A note on privacy
A text-to-speech engine receives every piece of text that apps send it to speak. With TalkBack, that can include messages, emails and anything else on screen. Install engines only from Google Play or F-Droid, prefer developers you recognize, and check whether a voice needs an internet connection: Android’s API lets an engine mark voices that send text to a server. Offline engines such as Google’s downloaded voices, SherpaTTS, RHVoice and eSpeak NG process text on the phone. For documents with sensitive content, is text-to-speech private? has a checklist.
When an app is the better answer
If you mostly want to listen to long reading, such as PDFs, ebooks, articles and study material, a better system voice only gets you part of the way. Select to Speak reads a screen at a time, and screen readers are built for navigation rather than chapters.
A dedicated reader app keeps your place, plays in the background with the screen off and handles whole files. Frateca runs on Android, iPhone, the web and Chrome with one library and natural voices, and the free plan has a daily allowance with no card. You can hear its voices in the live demo before installing anything. For listening without a connection, offline text-to-speech covers which apps and engines work in airplane mode.
Stop reading. Start listening.
Frateca turns PDFs, articles, textbooks and web pages into natural audio you can play anywhere: on your commute, at the gym, or while you cook. Free plan included, no card required.
iOS · Android · Web · Chrome extension · Free plan, no credit card required