Find a replacement for freetts

[Pierre Labastie]

Freetts has not been updated since 2009. It is written in Java, and does not run with Java 9. It has been archived for version 8.2. But a speech generation system is nice to have, so we should seek a replacement. See also #7754.

[Pierre Labastie]

Some of the possible replacements (please feel free to update):

• I've found those among the modules of ​Speech Dispatcher:
  • ​CMU flite
  • ​CMU festival
  • ​espeak

• Lot of people seem to like ​SVox PicoTTS. It's now developped by google. But the build system is android, and I'm not sure it is easy to use in LFS. Debian provide a package, so it is possible to build from source (I may dig into that later).

[Bruce Dubbs]

It may be more than TTS, but should we consider Gnome Orca (​https://help.gnome.org/users/orca/stable/)?

[Pierre Labastie]

I think Gnome Orca by itself does not do tts (and actually, it can be set up to not do tts at all, but rather to output to braille). It has to be set to use a tts program. But it may be taken in consideration which ones can be used.

A quick browsing shows that orca is built upon speech-dispatcher (for voice output), which itself can use, at least, espeak, flite, or festival (they seem to recommend the last one in documentation) for tts.

Note that both orca and speech-dispatcher have dependencies that are not in the book at the moment.

[Douglas R. Reno]

I think adding Orca is our best bet here. If we still don't have any new tickets by Wednesday, I'll add Orca, espeak, and speech dispatcher. It'll also silence a warning in gnome-shell, but I need someone on sysv to test the changes when I've done them, if possible.

[Douglas R. Reno]

I'll reply here with an outline of dependencies required in a bit, because there are many of them.

[Douglas R. Reno]

To start off with, dependencies which are present in the book already:

• at-spi2-atk
• gsettings-desktop-schemas
• gst-plugins-good/base
• gtk3
• pyatspi
• dbus-python
• pyxdg
• itstool
• git (optional)

However, now the complexity begins. Per Arch, the packages we don’t have in the book right now are:

• brltty
• liblouis
• speech-dispatcher

The dependencies for those are plentiful. Here’s a list of dependencies that we’ll have to add (in total). I’ll note here which packages can have external dependencies converted to internal, and those will have to be tested.

• Brltty

espeak

◦ Portaudio (mpg123, wireshark)

◦ JACK (OpenCV, xine-lib, alsa-tools, mlt, alsa-plugins, pulseaudio, mplayer, vlc, mpg123, audacious)

◦ celt (ffmpeg)

◦ zita-alsa-pcmi

◦ zita-resampler

• Speech Dispatcher

◦ dotconf

◦ libspeechd

◦ espeak-ng

◦ pcaudiolib

• liblouis

That’s 13 packages that would have to be added, which is large especially considering the amount of testing required. I can still do it if wanted, but I think it would be better to wait until after 8.4 is over.

[Bruce Dubbs]

I don't know if we want to support this at all. It seems like an awful lot to support one package. It's more effort than it's worth.

[Pierre Labastie]

Qt5TextToSpeech can use flite and/or speech dispatcher

[Berzerkula]

This isn't a new enhancement, and freetts has been gone a long time ago. You won't find a replacement. I suggest closing this ticket.

[pierre]

I think the way to go would be to add orca and its dependencies to the book. This would be great for eye impaired people, I guess, but by lack of manpower, it cannot be done ATM.