A small, fast speech engine for Windows and the command line
eSpeak NG is a speech synthesizer: it converts written text into spoken audio. It is a continuation of the older eSpeak engine, kept deliberately small so that it runs on modest hardware and responds instantly, and it covers a very large number of languages and accents. On Windows it installs both a command-line program and a system voice that other software can select, which is why it appears inside screen readers, learning tools and accessibility setups rather than only as a standalone application.
What the program does
The engine takes text and produces sound, either through the speakers or into a wave file, with options for voice, speed, pitch, amplitude and word gap. Its language coverage is the distinctive feature: major languages plus regional variants, minority languages and constructed ones, each with its own pronunciation rules. The Windows package registers a voice with the system speech interface so that programs which use the operating system's text-to-speech can choose eSpeak NG, and it also provides the command-line tool for scripting and batch conversion.
The working rhythm
Most people install it once and then select the voice inside whatever program actually needs to talk: a screen reader, a reading tool, a language exercise or an automated telephone-style prompt. From the command line, a short invocation reads text, and adding a flag writes the audio to a file instead, which is how narration for a teaching resource or a batch of pronunciations gets generated. There is no editor and no library to manage, so the workflow is whatever the surrounding application imposes.
Settings and limits
The sound is formant synthesis, which is intelligible and extremely lightweight but clearly robotic compared with modern neural voices; users who want natural narration will prefer a different engine. Some languages are implemented as accents of a related language rather than as full independent voices, and a few rely on external diphone databases that must be downloaded and licensed separately. The Windows package is a system component rather than an application, so there is no window to configure: settings live in command-line options or in the host program's voice list.
Who should choose something else
Anyone producing a podcast, audiobook or voice-over will get far better results from a neural text-to-speech service, and the voices already included with Windows sound more natural for everyday reading. People who need text read from the screen want a screen reader rather than a raw synthesizer. eSpeak NG is the right choice when the requirement is broad language coverage, tiny size, scriptability and no cost.
- Best for
- Users and developers who need a lightweight, scriptable text-to-speech engine on Windows with coverage of many languages and accents.
- Good to know
- The voice is synthetic rather than natural, there is no graphical interface, and a few languages need separately licensed diphone data to work.
How to get started
- Download the eSpeak NG MSI from the project's release page and install it.
- Open a command prompt and run the espeak-ng command with a short phrase to hear the default voice.
- List the available voices and pick one with the voice option, for example a different language or accent.
- Adjust speed, pitch and amplitude until the output is comfortable, then save the settings in your script.
- Write the same text to a wave file if you need audio for another program or a lesson.
Questions & answers
Where will the installed voice appear?
The MSI registers a system voice, so programs that use the Windows speech interface can select eSpeak NG from their voice list, including screen readers and reading tools.
Why does it sound robotic?
It uses formant synthesis rather than recorded or neural speech. That keeps it small, fast and usable on old hardware, at the cost of naturalness.
Can it read text in languages Windows does not have?
Often yes. eSpeak NG includes many languages and regional accents, although a few depend on additional diphone data that must be installed separately.
