How does eleven labs generate the speech? Can you specify one speaker for an NPC id?

You create a voice called race-gender and the cli picks it up given your API key. Currently there is 1 model per race-gender combo but some people in the discord have worked on npc specific models like thrall.