Blather : speech recognizer


Blather is a speech recognizer that will run commands when a user speaks preset sentences.


  1. pocketsphinx
  2. gstreamer (and what ever plugin has pocket sphinx support)
  3. pyside (only required for the Qt based UI)
  4. pygtk (only required for the Gtk based UI)


  1. move commands.tmp to ~/.config/blather/commands and fill the file with sentences and command to run
  2. Run, this will generate ~/.config/blather/sentences.corpus based on sentences in the 'commands' file
  3. quit blather (there is a good chance it will just segfault)
  4. go to and upload the sentences.corpus file
  5. download the resulting XXXX.lm file to the ~/.config/blather/language directory and rename to file to 'lm'
  6. download the resulting XXXX.dic file to the ~/.config/blather/language directory and rename to file to 'dic'
  7. run

    • for Qt GUI, run -i q
    • for Gtk GUI, run -i g
    • to start a UI in 'continuous' listen mode, use the -c flag
  8. start talking


once the sentences.corpus file has been created, run the script to automate the process of creating and downloading language files.

To run blather with the GTK UI and start in continuous listen mode:
./ -i g -c