Voice sounds mechanism (Patreon)
Published:
2019-11-24 14:01:01
Edited:
2019-11-26 01:20:00
Imported:
2021-09
Content
I will explain the improved voice sounds mechanism based on the reflection of the previous work.
- 1. When the action is executed. randomly selected one from multiple voice candidates. (ex. "Uh!", "iyaa", "kya!", "dame")
- 2. Suppose "iyaa" is chosen. Determine whether to use Japanese or English according to "Speaking language ratio" setting or when "Natural bilingualism" is on, It follows the specific ratio set for each voice. If English is selected, it's assigned to "Noh".
- 3. And further, there are several "Noh" sounds, so select one of them. Suppose "Noh03" is chosen.
- 4. Up to 6 voice histories output in the past are stored, so it will be output if there is no same voice "Noh03" in them.
- 5. If there is the same voice in the history, start over from 1. There is almost no overlap at this point, but if a different voice is not found after repeating a certain number of times, delete the history references one by one from the newest one and repeat more.
- 6. Final selected voice sound is output and record it in the history.
In this way, the same voice is not played continuously as much as possible. In the case of speech, in addition to this, there are multiple tags for each line, and filtering is performed according to the state of clothing etc, although it doesn't have so many voices.