Upload as Text should Transcribe audio attachments as Text

[RolRodr]

For models that don't allow direct audio uploads, it would be nice if users could upload an audio file and use the SST model they have configured to pass in the text from the audio into the prompt.

I'm not sure if I'm missing some other configuration that allows for uploading audio files generally for all models. Right now, audio upload only works for models from the Google endpoint.

Thank you!