How to control the details of AI vocal performance for better results?
In ACE Studio, you can control the details of the AI vocal performance by adjusting the consonant length, pitch, and AI parameters.
In ACE Studio, you can control the details of the AI vocal performance by adjusting the consonant length, pitch, and AI parameters.
Consonant length
After you click the play button and the singer track completes the AI rendering, you'll notice that some notes in the clip view have semi-transparent sections at the beginning and end. These represent the consonants of the lyrics corresponding to those notes. Conversely, the solid parts represent the vowels of the lyrics for those notes.
You can drag the boundaries of the consonants to adjust their length. Please note that the pre-consonant is positioned outside the note duration (before the note's start time), while the post-consonant is within the note duration (before the note's end time).
Pitch
You can use the "Draw Pitch" tool to draw pitch lines and the "Pitch Eraser" tool to erase the drawn pitch lines, restoring them to their default generated state. Use the "Modulation" tool to drag up and down on a note to adjust the pitch smoothness, and the "Vibrato" tool to click on a note and drag to add vibrato. The AI singer will perform according to your modified pitch.
AI parameters
By adjusting the four-dimensional emotion parameters, you can change the expressiveness of the AI singer's performance.
Four-dimensional Emotion Parameters
- Air: the amount of breath during singing
- Falsetto: the amount of falsetto during singing
- Tension: the degree of tension/relaxation of the vocal cords during singing
- Energy: the energy and strength of singing
This example demonstrates how to make the AI vocal sound more gentle or aggressive by tweaking these AI parameters.
However, please note that the impact of these AI parameters on the singing performance is relatively opaque, unquantifiable, and varies for each AI voice. Therefore, it's important to first select the AI voice you want to use, render the AI performance, and listen to the results. Then, adjust the parts you are not satisfied with, rather than making large-scale, blind modifications to the parameters.