Add speech endpoint

[m1guelpf]

• Added a post_raw and execute_raw internal functions, since we don't want to JSON-decode the response body for the speech endpoints (which return raw audio). Modified the execute function to wrap around execute_raw.
• Added basic support for the TTS api

Note that the struct corresponding to the response_format parameter was named SpeechResponseFormat instead of AudioResponseFormat, since the later already existed for defining whisper response formats.

I'd love to add an enum for voices as well (with an Other wildcard and marked as non-exhaustive), but refrained since the library's pattern seems to be deferring to strings (same case as model IDs all across).

[64bit]

Thank you! 🎉

( There's https://crates.io/crates/serde_bytes too, not sure what difference would be between bytes crate and serde_bytes )

'd love to add an enum for voices as well (with an Other wildcard and marked as non-exhaustive), but refrained since the library's pattern seems to be deferring to strings (same case as model IDs all across).

I think for Voice as enum makes sense - go for it . I usually defer to spec when it lists a filed of type string + enum. They haven't released updated spec though.

---

I'm getting "The model tts-1 does not exist or you do not have access to it." Not sure how to get access to the new models?

[64bit]

OH they just did release spec 11 mins ago :D

[m1guelpf]

Not sure what the difference is either, but bytes is the type reqwest uses, we just import the crate to access the type really.

Will add the enum in a sec. Model releases in 1h (supposedly) so not available yet lmeow

[64bit]

Aha here, https://openai.com/blog/new-models-and-developer-products-announced-at-devday

We’ll begin rolling out new features to OpenAI customers starting at 1pm PT today.

30mins to go 🕐