Allow custom Voice and Speech Models

[adrianliechti]

this pull request removes the client-side logic to verify a valid speech and voice model.
the motivation is to support custom models on openai compatible backends

thanks for considering

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 98.66%. Comparing base (774fc9d) to head (f1994e2).
Report is 16 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #691      +/-   ##
==========================================
+ Coverage   98.46%   98.66%   +0.20%     
==========================================
  Files          24       24              
  Lines        1364     1125     -239     
==========================================
- Hits         1343     1110     -233     
+ Misses         15        9       -6     
  Partials        6        6              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[sashabaranov]

@adrianliechti thank you for the PR! I'm curious — are there any backends now that do support those?

[adrianliechti]

@sashabaranov thanks a ton for coming back on this!
my own one adapts tools like mimic3, coqui.ai, whisper.cpp compatible ones into an openai-compatible api (also multiple ones). here a link: https://github.com/adrianliechti/llama. It is in use as part of a commercial platform at my employer.
and since I was using your lovely SDK for both tests but also integration, I allowed me to create this PR.

I think it is in line with this project's recent changes on embedding/models - so I hope over time a broader audience benefits (or at least needs less urgent maintenance once openai releases new whisper or tts models)