Upstream contribution · #823
Honor response_format on /v1/audio/speech
Not adopted. Three months later the maintainer solved the same problem differently, avoiding an ffmpeg dependency.
This pull request was not adopted. It is recorded because an account of the work should include the parts that did not land.
The problem was real: the endpoint accepted response_format and returned WAV regardless. My implementation brought in ffmpeg to transcode. Three months later the maintainer solved the same problem along a route that added no system dependency.
The problem got fixed; my code did not. I have no evidence that this pull request influenced the eventual implementation, so I do not claim it did.