Changing the default text-to-speech voice

Tell application "System Preferences" 
    set "default voice" to "Agnes" 
end tell 

The result is:

Can’t set "default voice" to "anna". Access not allowed.

There are two problems with your approach:

• The dictionary for the System Preferences app contains no default voice element or any other for changing the TTS (text-to-speech) default voice (as of OS X 10.11); in fact, it seems that Apple provides no programmatic way of changing the default voice (not even via its NSSpeechSynthesizer Cocoa class).
• By double-quoting default voice, you're trying to assign a value to a string literal, which will always fail.

^{Note: An earlier version of this answer pointed to a Bash script named voice in a Dropbox location; this script has since been renamed to voices, had its syntax revised, and is now properly published as an open-source project - see below.}

Unfortunately, as of OSX 10.11 (El Capitan), there is no documented programmatic way to change the default voice.

It can be done, but doing so requires on undocumented system internals, so future compatibility is not guaranteed.

voices is a CLI I wrote that does just that - verified to work on OSX 10.11 down to OSX 10.8.

You could then do the following from AppleScript :

do shell script "/path/to/voices -d {voiceName}"

For instance, if you place voices in /usr/local/bin and want to switch to Agnes as the default voice, use:

do shell script "/usr/local/bin/voices -d Agnes"

If you happen to have Node.js installed, you can install voices to /usr/local/bin with

npm install voices -g

Otherwise, follow the instructions here.

Changes to ~/Library/Preferences/com.apple.speech.voice.prefs.plist seem to be applied immediately.

d=com.apple.speech.voice.prefs
if [[ $(defaults read $d SelectedVoiceName) = Kathy ]]; then
  defaults write $d SelectedVoiceCreator -int 1835364215
  defaults write $d SelectedVoiceID -int 201
  defaults write $d SelectedVoiceName Alex  
else
  defaults write $d SelectedVoiceCreator -int 1836346163
  defaults write $d SelectedVoiceID -int 2
  defaults write $d SelectedVoiceName Kathy
fi

Another option using UI scripting:

tell application "System Preferences"
    reveal anchor "TTS" of pane "com.apple.preference.speech"
end tell
tell application "System Events" to tell process "System Preferences"
    tell pop up button 1 of tab group 1 of window 1
        delay 0.1
        click
        if value is "Alex" then
            click menu item "Kathy" of menu 1
        else
            click menu item "Alex" of menu 1
        end if
    end tell
end tell
quit application "System Preferences"

Without the delay the value was Loading Voices… if System Preferences wasn't open before.

To get it working with Yosemite, you will need to add the following 2 lines to the bottom of the script provided by mklement0 above:

Original Link to the file from mklement0: https://dl.dropboxusercontent.com/u/10047483/voice

Add the two lines below to restart SpeechSynthesisServer, otherwise you can't use the shortcut key to immediately access the new default voice:

killall SpeechSynthesisServer
open /System/Library/Frameworks/ApplicationServices.framework/Versions/A/Frameworks/SpeechSynthesis.framework/Versions/A/SpeechSynthesisServer.app

This is working:

property currentVoice : "Vicki"

set systemVoices to {"Agnes", "Albert", "Alex", "BadNews", "Bahh", "Bells", "Boing", "Bruce", ¬
 "Bubbles", "Cellos", "Deranged", "Fred", "GoodNews", "Hysterical", "Junior", "Kathy", ¬
 "Organ", "Princess", "Ralph", "Trinoids", "Vicki", "Victoria", "Whisper", "Zarvox"}

    repeat

     activate me

     set theResult to display dialog "Say What?" default answer ¬

      "" buttons {"Quit", "Speak", "Change Voice"} ¬

      default button "Speak" cancel button "Quit"


if button returned of theResult is "Quit" then exit repeat

 else if button returned of theResult is "Change Voice" then

  set currentVoice to item 1 of ¬

(choose from list systemVoices with prompt "Choose new voice.")

 end if

 if text returned of theResult is not "" then

  say text returned of theResult using currentVoice volume 1

 end if

end repeat