I love making notes. I’ve written about it a little on this site: Matt GTD.
Sometimes it’s not feasible or it’s inconvenient to write a note, so I record the idea on my phone with a voice recording app. I’m currently using Easy Voice Recorder.
Easy Voice Recorder has excellent cloud storage integration options. I was using those, as I had an affinity for cloud related services. However, with Security implications, along with Evernote losing my data, I’ve been more incline to move things “in house” and do things myself.
Syncing #
The first thing I had to do was get the voice recordings off of my phone and into a central computer. I use Syncthing for that. Syncthing is an open source project that takes security very seriously. “Out of the box” it works very well. If you are more security conscious, you can lock it down further with things like disabling NAT traversal, Global Discovery and/or even running your own Discover Server.
As a side note, it’s pretty kick ass having a two way sync set up on my phone. Sure, having the data flowing from my device to a server is awesome, but dropping a ringtone or whatever in a folder on my computer to send to my phone is really convenient.
Transcription #
Once I had the voice recordings syncing to a server, I decided to investigating Speech to Text for the audio files to save me from transcribing note manually. After some investigation I found the Faster-Whisper python library.
The focus for this library is to use large models on GPUs to do very fast transcribing. I don’t need fast transcriptions. I batch them overnight. For my purposes, the small model on CPU works perfectly fine.
In the snippet below, you can see that the code walks the recordings directory, looks for a corresponding text file and does the transcription if no text file exists.
Summary #
Essentially, that is it.
tl;dr #
(Easy Voice Recorder) -> (Syncthing) -> (Faster Whisper)
I put the script in a cron job to run every night. As I mentioned at the top, Faster-Whisper works very well. So well, that one time I sang a melody into a recording to remember it and it accurately transcribed my “D-D-D-D"s! 😂
All of that said, whether or not you want to undertake the transcription process, I highly recommend setting up Syncthing. Having the ability to sync your data between ALL of your computers and phones, for free, without the cloud, is a win.