okay got it working, using minimax 2.6 turbo at the moment. sounds way better than turbo 2.0. However, im getting really bad latency. Passing my certificates with Lets Encrypt integrated into my node.js server so there shouldnt be any super long dns latency, im seeing it more come from Minimax even though their new models supposed to be <250ms, im aware itll be more feeding through to vapi, but im getting like 800-900ms on average to get the audio response to Vapi which is significantly higher than older natively integrated models. I found that the culprit is native Vapi can stream tokens and chunks of text and audio, while it seems my current server is not (waits for the entire TTS response before sending it all as one chunk), I was wondering if thats implementable with custom tts since Minimax does support that. Also, if possible could the UI include a custom voice option since even though i can edit it with the API, once i try changing any other assistant settings in the UI it wont let me publish it since it thinks the voice json information isnt acceptable