Any solutions for Proscan to readable text alerts?

[tomas21]

Proscan is an amazing product! Curious if anyone has come up with a solution to process the recordings or live audio into readable text alerts?
I'm sure this is what the Citizen app platform does but would be nice to have it personally ... with the advancement of opensource and AI seems more feasible these days..

 

[MaineGuy]

No SMS alerts, but readable text and summaries. A bit of work to set up, but then it works pretty well.

Whisper speech to text is acceptable. Mostly. Sometimes comically bad!

I'm sending audio files from ProScan and sdrtunk to it. I run it 24/7.

 

[tomas21]

Sweet so your using the container image of scanscribe to solve this?

 

[MaineGuy]

The whole Docker thing was new to me.
I just followed the instructions as best I could.

I turned off the Ollama stuff - I couldn't get it to work ( PC too slow? ). I let Gemini do the hourly summaries.

GitHub - xxbubziexx/ScanScribe: ScanScribe is a powerful AI powered open source public safety radio transcription suite built for serious hobbyists.

ScanScribe is a powerful AI powered open source public safety radio transcription suite built for serious hobbyists. - xxbubziexx/ScanScribe

github.com

 

[tomas21]

Awesome.. giving it a go... getting all the prereqs, .env, config.yml, and all the settings sorted out.

 

[tomas21]

Very cool solution. Need to figure out filter out recordings before they hit the ingest folder and looking for keywords unique to our area

 

[MaineGuy]

I don't think you need to filter out recordings before ScanScribe gets them.
You can set minimum size or duration for processing. ScanScribe then uses voice detection to remove other junk.
Processing requirements aren't so bad that I felt I needed to do more.
Good luck!