Very cool! I am curious about the technical limitations of the duration of the audio files. Any way to overcome that? (Besides internally splitting them)
I am also curious about that limitation.
I tried the lib with a 10 mins audio and it gave me some unrelated words but when I tried a short clip of 20 seconds, it was able transcribe it properly. I think I have to dig deeper to know what caused this behaviour.
It's due to the way we've defined the graph in Tensorflow. Static shapes allowed for faster training. We'll be working to remove this limitation in subsequent versions.
at16k is a Python library to perform automatic speech recognition or speech to text conversion. The goal of this project is to provide the community with a production quality speech-to-text library.