Any way to get the voices (Natalie in this case) to approximate talking, like sometimes happens in a song?
If you can input the correct timings and pitch changes, yes, however the software is not a speech synth and the AI engine has not been trained on any spoken learning material, so it relies entirely on you identifying the correct patterns and replicating that with notes and parameters.
Some of the AHS voice databases have a speech counterpart for different engines, but those are a separate product. For example:
- Koharu Rikka, Tsurumaki Maki, Natsuki Kari, and Hanakuma Chifuyu all have a “talk” version for CeVIO AI
- Kotonoha Akane/Aoi have a talk version for A.I. VOICE
Of the above options, the only English talk voice available is Tsurumaki Maki English for CeVIO AI Talk.
Dreamtonics has created a speech synth software called “Voicepeak”, however at the moment it only supports Japanese and there are no voice databases released for Voicepeak which also have a SynthV counterpart.
I was told by Audiologie that “There is a way to manipulate our voicebanks into talking through manual tuning , we will demonstrate this ‘talkloid’ method in our demo next week.” (private mail exchange)
Not sure what they will bring to the table, but could be interesting.