Menu
Inspired by a tweet from Kevin Regamey about how the voices of Celeste were made, I tried to emulate this same feature using UE5 newest features. So I did some research on the Fmod project that they have made public to see how the voices were actually made.
I learned about formants and how they can be synthesized, and created a little demo in the side that you can listen to. This demo changes and manipulates a synth over time to create speech-like patterns.
In addition to the sound system, I added the profile pictures to the dialogue boxes in order to strengthen the emotional intent of the dialogue.
Using the variables you can see to the left here, you can get a surprising amount of expression from the voices. And yes, you can even set a seed to make the whole system fully deterministic.
In case you don’t know what formants are, these are the frequencies that get boosted naturally by the human mouth and throat and shifting these up or down creates more feminine or masculine voice qualities. You can then combine those with Pitch to create a very convincing effect.
If you are interested in learning more, I can highly recommend Felix Burkhart and Walter F. Sendlmeier’s research paper: “Verification of Acoustical Correlates of Emotional Speech using Formant-Synthesis”