I recently ran across an interesting open source project called “Mico Project“. Originally, I was researching the LIUM Speaker Diarization Tools for use in some work projects and saw that the Mico Project utilized it.
What drew me to the Mico Project was it purported to do both speaker diarization and speech to text. It also had a neat twist of associating various types of media together, so not only could you convert some audio to text, but one could also associate it with other types of information. Very cool.
So with an end goal of diarizing and transcribing some audio, I set out to install a protoype and play with it. Unfortunately, in the end I could not get it working, but hope writing this blog post will attract some support.
Attempt 1 – Installing Mico Virtual Machine Image
Following their guide posted here https://www.mico-project.eu/pages/documentation/#document-11 I was able to download and get the virtual machine running. I was able to execute a workflow. The problem is the virtual machine image does not contain the diarization or transcription extractors out of the box. I found information on installing extractors here, but I cannot find the root password to install them on the virtual machine and the one user they provide login information for, mico, does not have sudo privileges. So I cannot find a way to install additional extractors to the virtual machine image. Bummer. However they do provide instructions on doing your own install on a Debian box. Let’s try that!