More moods/emotions/vocal modes, and how to get them

DDBE

• January 08, 2023 10:29

Request (if it's not already implemented): A way to automatically identify (both in training of a voice model as well as in material that needs to be converted) of the seven basic modes of articulation or phonation, in order for the app to maintain them in the conversion: https://en.wikipedia.org/wiki/Phonation#State_of_the_glottis

The way how to get there could benefit from some open-source code by the public French IRCAM institute (one only needs to register a free account in order to download the code or plug-ins) that could be used by the app to automatically, internally modify the vocoder output according to the input once the model can recognize those features in the input:

- https://forum.ircam.fr/projects/detail/angus/ (used to turn any voice recording into various degrees of shouting or yelling)

- https://forum.ircam.fr/projects/detail/david/ (used to change, switch, or add various emotions to the sound of a voice)

- https://forum.ircam.fr/projects/detail/vibra-voice/ (used to add or modify vibrato on a singing voice)

5

• 

  BrightShadow

  • February 10, 2023 01:39

  Anything that further improves the range of this would be great!

  2
• 

  JonBap the Bold

  • March 16, 2023 17:07

  This should be a high priority, it will make voice.ai far more useful when the software can either detect when you are shouting, angry, etc. or if it had toggles to change the emotion of the voice output.  Right now it is great if you are trying to create a calm speaking voice, but as soon as you try to act and put some emotion into it, the results come out flat.  You could angrily shout out "You killed her!" and get a comically relaxed "You killed her."

  I would think this is within the realm of possibility, you would just need samples of voices in various states of emotion for the AI to train on.

  What the Voice.AI team has created here so far is extremely impressive, my hat's off to them.  I am just hoping to see emotional range implemented so that I can use it in voiceover projects that require more than narration style talking.

  3

Please sign in to leave a comment.