My question is if you have to convert only one recording to text then both would be okay, but in case there are hundreds of recording which are to be transcripted simultaneously then which would be a better option and will not have delay in the response?

I have tried using both for a single file but the one using s3 is a bit complicated as you have to store the file into the s3 and then convert it to text, but the other way i think is if the real time transcription api is bombarded with 100s of audio recordings at once will it still be the same without any delay?