Could you elaborate a bit on "my speech engine tied to a specific model provider"? Sorry, I might be lacking some context on what you are referring to here.
I will be in the market for a text-to-speech engine, but from looking at the website it seems the model of Retell is trying to push is "use our all in one model + text to speech service" which is problematic when my choice of model and control over how that model runs is at the core of my product, and text to speech is a "nice to have" feature. I want an endpoint that I can fire off text to in a streaming mode, where it'll buffer that streaming text a little and then stream out a beautiful, natural sounding voice with appropriate emotion and intonation in a voice of my design. I'm sure I'm not really Retell's ideal customer, and they're going after lucrative "all in one" customers that just want to build on top of a batteries-included product.
If you are looking for a text to speech solution, you could use elevenlabs turbo model.