model with many voices, what is the point ?

[mishal-harbi]

Hi,
I'm new to Ai and learning along the way. I am just wondering why the researchers didn't just train a single model for each voice (male or female) to produce a more efficient and highly tuned model? I ran a very short file (100 words?) through Tortoise and it used up all the 24 GB of my 4090 with the Task Manager showing GPU Memory 38.8/55.9 GB. Also the voice segments produced each has a different voice. Also it took a long time to complete the job.
just wondering about the advantage of using this approach.