Hello everyone,
I was testing two different endpoints to speech-to-text task, one of them is “whisper-large-v3”
I was testing the output for an empty audio, the other endpoint is giving me an empty string, which is the correct output, but whisper is giving me this output: " you" ! No matter how long the file is, from 1 second to 20, it gives the same answer. Is this the expected output for empty audios? I didn’t find this information anywhere, also did someone try before to test it against only noise background without any talking?
I tried an audio with bird singing and it gave me this output “Thank you”
Thank you!