4. Problems

There are many known problems with Speakup/SpeechD-Up. You might want to consider using some user-space assistive technologies if you want to use software synthesis on the console. Such technologies are speechd-el, Emacspeak, SuSE Blinux and Yasr. We however believe the Speakup/SpeechD-Up combination is useful even though it has the described deficiencies.

• (Fixed: Speaking characters and keys is slow)

  The original issue was fixed (but see bellow). If you are still not happy with the performance while reading characters and typing in, then you are using a synthesizer which is both slow and lacking support for caching (use Festival, it does caching with Speech Dispatcher and keys reading is instant).

• Characters behave like KEYS for synthesizer configuration

  The way Speakup passes individual characters makes it impossible to tell whether this character originated as a key or while the user was moving the cursor on the screen and the already present letters are being read. We believe it is important to carry the proper information when pressing keys (like capitalization), so SpeechD-Up handles these messages as keys.

  Also space, tab and various other whitespace characters are not reported as a characters/keys as this is impossible to achieve with current Speakups implementation. It is impossible if the word “space” in input originated by the user pressing the SPACE key or by encountering this word inside a document being read on the screen.

• (Fixed: It doesn't speak my language / it uses some weird character set)

  If you still experience this issue, you are probably doing one of this

  • Using a wide character encoding like UTF-8. This is not supported for inherent issues with wide characters inside the kernel.
  • Not running speechd-up with the -c parameter and specifying the proper character set your console uses or not setting the SpeakupCoding option in the configuration file. Alternatively, you may not be specifying the language to use. You can choose a language either in Speech Dispatcher configuration or via --language or in the ‘speechd-up.conf’ configuration file.
  • Running speechd-up with the -t or --dont-init-tables. This option is incompatible with internationalization.

  If this is not true, please contact us on speechd@bugs.freebsoft.org.

• The spelling/punctuation modes of my favorite synthesizer don't work.

  Speakup uses it's own very simple spelling and punctuation substitution and doesn't allow the TTS to use it's own user-specified rules.

• It doesn't work in UTF-8

  UTF-8 should work as long as you only use characters encountered in the basic ASCII. So English should work. But Speakup doesn't really support UTF-8 in any way (nor it can, there is no UTF-8 support inside the kernel – just workarounds targeted at visual interpretation; the kernel virtual terminal drivers contain many barriers for accessibility).

• Punctuation and capital letters recognition doesn't properly work

  This is an issue again related to Speakups design. It is currently being worked on.

• I can't switch languages or synthesizers on fly

  The way the interface Speakup – Speech Dispatcher is designed, this is not possible. You can switch them statically in Speech Dispatcher's configuration though. If you use SpeakUp's identification "*:speechd-up:*" you can even choose a synthesizer specifically for Speakup.

• Any other more advanced thing doesn't work or something is too slow or ... There are many problems for software speech with how this interface is done and with how Speakup is designed. It's probably still better to use Speakup with a hardware synthesizer, if you have the possibility. Speakup was not designed with user space software synthesis in mind and the necessary changes would probably be too difficult now. There were many important design decisions made which are very reasonable for hardware synthesizers, but they just don't mix well with software synthesis.

  On the other hand, experience shows people still find this interface to software speech synthesis useful, mostly because of lack of anything else. Also, it seems the future of accessibility technologies lies in the software synthesis (extensible, etc.) as machines get faster. That's the reason why we provide this application and we will try to help you if you have problems or questions.

This document was generated by Hynek Hanke on March, 26 2009 using texi2html 1.78.