Emacsvox retains direct speech servers for existing installations. They are not the recommended first setup and are outside the initial-user platform matrix above:
espeakThe direct GNU/Linux Tcl server. Building it with make espeak
requires Tclx at runtime, Tcl 8.6 development headers, eSpeak NG development
libraries, and a C++ compiler. It remains useful for compatibility testing,
but new users should install Omnivox instead.
swiftmacThe direct Swift macOS server. It requires Swift 5.9 or newer and is built
with make swiftmac. It remains a compatibility alternative to
Omnivox’s AVSpeechSynthesizer engine; see
servers/mac-swiftmac/Readme.org.
macThe legacy Python/PyObjC macOS server. The unset runtime selector still names it as a compatibility default on Darwin, but it is not the recommended or documented initial-user route. It depends on Foundation, AppKit, and PyObjCTools availability.
outloudThe direct GNU/Linux IBM TTS or Eloquence adapter. It requires a separately licensed compatible runtime, which Emacsvox does not distribute. Voxin availability is documented at https://voxin.oralux.net.
dtk-softThe direct Software DECtalk adapter. Install its runtime separately and see
servers/software-dtk/Readme.org. Emacsvox does not distribute licensed
speech runtimes.
The native Windows audio, standalone Eloquence, and standalone DECtalk bridges
under servers/ are advanced WSL-specific compatibility paths. Piper is
an optional Omnivox engine and model workflow, not a primary Emacsvox server.