I've received a number of queries asking about the differences between Emacspeak and Speechdel ---especially given some of the somewhat confusing assertions made in recent Speechdel release announcements. I'm posting this article in the hope of clearing up some of this confusion.
Emacspeak speech-enables Emacs by advising core emacs functionality. Speech services are provided by a simple Emacspeak speech-server. Additionally, Emacspeak implements speech-extensions for popular emacs modules --- see the speech-enabled applications list.
Emacspeak was first released in 1995, and then (as now), there was limited speech access to the Linux GUI. Therefore, to be useful as a complete access solution, Emacspeak has always needed to enable the user to do everything from within Emacs, not just regular editing operations. As a case in point, emacspeak users are probably one of the last remaining communities that use Emacs for browsing the Web.
The idea of SpeechD --- peech Dispatcher as an intermediate layer between speech clients and TTS engines was first floated sometime in the late 90's. Such a common layer is a laudible goal but is something that takes time and effort to get right. Additionally, you have the challenge of geting existing software e.g., emacspeak, to abandon their own speech abstraction and re-implement against a supposedly more generic, but completely untested and untried intermediate layer.
The developers of SpeechD initially incorporated some of the Emacspeak code into an Emacs wrapper (speechdel) that called SpeechD, but later decided to go their own way -- and present speechdel is the result.
Like Emacspeak, speechdel uses Emacs Lisp's advice facility to add spoken feedback to core editing commands; speech output is produced by calling out to speech-dispatcher.
The summary of feature differences between Emacspeak and speechdel in the next section is from examining the speechdel code-base; I have not run speechdel since its dependency chain resulting from speechd was difficult to resolve on my FC3 64bit machine.
Emacspeak implements Aural CSS ACSS, and uses it to provides
the aural analog of
- Emacspeak provides pronunciation dictionaries. Pronunciations can be defined on a per-mode, per-buffer or per-directory basis. Directory and mode specific pronunciations are persisted across sessions. This allows Emacspeak to leverage Emacs' intelligence about the semantics of a given application; thus, you can have it say "p arrow x" for "p->x" when editing C code. Per-directory pronunciations are useful for reading electronic books. Per-buffer pronunciations are useful for succinctly speaking long lines of shell output e.g. when compiling complex software.
By advising core Emacs functionality, Emacs modes work out
of the box with Emacspeak. But in most cases, Emacspeak goes
one step further by providing light-weight speech-modules
that specialize spoken output for a given mode. As an
example, advising next-line to speak the current line is
sufficient to use
dired-mode--- but having to listen to the entire line of dired output is not a pleasant experience. The dired-specific module in Emacspeak advises all interactive dired commands to speak the "right" information. As an another example, GUD interaction automatically speaks the line of source-code without leaving the Gud buffer.
- Emacspeak comes with many "Emacs Applets" for performing tasks that most users would perform outside of Emacs. Examples include playing CDs, playing multimedia streams etc. Fortunately, I have not had to write too many of these since there are always Emacs users other than myself who also create such Emacs applications --- so where Emacs applications already exist, I merely speech-enable them with a small set of advice definitions, and in some cases add a few additional interactive commands.
Emacs applications are plentiful for most tasks; one
exception is the Web. Since emacs/w3 development was
abandoned sometime around 1998, I have added significant Web
interaction functionality to Emacspeak using Emacs/W3 as the
basis. Today, a lot of this has also been ported to
Emacs/W3M thanks to other enthusiasts on the Emacspeak
mailing list. Examples include:
- WebSearch module --- prompts for query and processes response to focus on the results.
- XSLT pre-processing: Allows pre-processing of complex pages before rendering via W3. Used to enable smart screen-scrapers using XPath.
URL-Tempaltes: Originally motivated by
webjump.el, this provides url templates that enable easy access to a variety of Web tasks rangig from looking up flight times to listening to your favorite NPR or BBc show. Think early cut at a "Web Command Line in the minibuffer".
Customization via Custom, including additional
keymaps. Comes with additional keymap files for the Linux
console to enable
- Module emacspeak-wizards iplements a large collection of Emacs wizards that enable common tasks that you would otherwise perform at the shell e.g., checking display status on a laptop. The additional prefix keymaps come in handy!
byte-compile-dynamicset to =T= and individual application-specific modules are kept completely independent of one another. Given the size of the Emacspeak codebase, this is a pre-requisite for both efficiency and developer sanity.