Skip to content
Branch: master
Go to file

Latest commit


Failed to load latest commit information.
Latest commit message
Commit time

Vox Ex Machina

Speaker verification voice-print security assessment tool

Features in vem.scd:

  • arbitary text to speech
  • real time speech parameter randomisation
  • enrollment and authorisation
  • replay and brute force attacks
  • add noise to the audio samples
  • API support for MS Azure Cognitive Services
  • API suport for VoiceIT

Additional features in development vemfuzzer.scd:

  • generate DTMF for fuzzing of interactive phone systems
  • generate SS5 tones for fuzzing of really old school phone networks
  • generate BlueVox tones for fuzzing of one time auio password systems
  • loading of a script for auto generation of audio
  • microphone access for capturing speech recognition system responses while fuzzing

Open vem.scd / vemfuzz.scd in SuperCollider and evaluate (select all and press "cmd+Return")

MacOS and Linux with these dependencies:

Some changes have been made to SCMage and the files are included here so you can replace the originals. There is also a precompiled compiled on Ubuntu 16.10 x86_64
There are currently some paths in a few files that need to be edited in the code. They are marked with "TODO:"

To get it all working:

  • build Festival and the SpeechTools using the scripts provided in SCMage. You must use gcc-4.8 otherwise you will get an "SIOD ERROR: unbound variable : eof" when labelizing the text files.
  • edit the file paths in and copy the relevant binaries from examples to bin dirs.
  • copy to ~/.local/share/SuperCollider/Extensions/classes.
  • copy to ~/.local/share/SuperCollider/Extensions/plugins.
  • make sure the data dir including here lives under the mage dir and supply a shortcut to it in the vox directory. The data files contain config with hard coded directories that must be edited to match your install.

The Festvox voices are under the directory "data".
Obviously you should have some experience using SuperCollider.

Basic help is available on the Wiki

A hardware implementation of this on Bela is in development. This will be all open source.


Speaker verification voice-print security assessment tool




No releases published
You can’t perform that action at this time.