lee/hpr-knowledge-base
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Initial commit: HPR Knowledge Base MCP Server

Browse Source 
- MCP server with stdio transport for local use
- Search episodes, transcripts, hosts, and series
- 4,511 episodes with metadata and transcripts
- Data loader with in-memory JSON storage

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
Lee Hanken
2025-10-26 10:54:13 +00:00
commit 7c8efd2228
4494 changed files with 1705541 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

102
hpr_transcripts/hpr3751.txt Normal file
View File

@@ -0,0 +1,102 @@
Episode: 3751
Title: HPR3751: Using Noisetorch
Source: https://hub.hackerpublicradio.org/ccdn.php?filename=/eps/hpr3751/hpr3751.mp3
Transcribed: 2025-10-25 04:55:50
---
This is Hacker Public Radio Episode 3751 from Monday the 19th of December 2022.
Today's show is entitled, Using Noise Torch.
It is hosted by Delta Ray and is about seven minutes long.
It carries a clean flag.
The summary is Noise Torch is a program for Linux that creates a virtual microphone
that removes background sounds.
Hello, I'm Delta Ray.
I wanted to talk about this new application that I found that apparently has been around
for a few years, but I found it after searching for information on how to make a podcast.
You might notice that I have some noise in the background.
One is a fan that is just running like a cooling fan that just blows breeze on me or whatever.
The other noise is, well, actually music, it's not noise, but it's background sound
that I don't want to necessarily get into a podcast, and that's my daughter playing piano
in the background.
It can be difficult to find time to actually do a podcast if you have to live in a house
with other people and they're doing their own thing or whatever, and you have maybe a noisy computer
and stuff.
There's this great program called Noise Torch.
I'll post a link to it in the show notes, but what it does is it apparently uses a
neural network to suppress noise and background noises other than like your own talking.
It's pretty easy to use, and it's written specifically for Linux.
Let me just give you a demo really quick.
Right now, I'm talking into my HyperX Quadcast-S microphone, and I've set Noise Torch
to be a virtual microphone, and I can actually switch over to it right now.
And you should notice that a lot of the background noises now dropped out.
Like I'll stop talking for a second there, and you probably don't even hear anything in the background.
And now I'll switch back to the normal microphone.
You can hear how much of a difference that makes.
And I'll switch back to the noise suppression.
Yeah, so the microphone's still running.
My daughter is still playing piano.
Excuse me.
I wonder if you heard me cough.
And it cuts out all of that.
So I can feel comfortable making like a podcast using this program without worrying about other background noises
getting into the mix.
You may live in a city or in a noisy neighborhood or something like that.
Maybe you live in an extra airport, and you're always having planes taking off or something like that.
This can enable you to make it easier for you to do podcasts.
So the way it works is you download this binary from GitHub.
They have releases, and then you run the program.
And it asks you what microphone you actually want to hook up.
So it will show you the microphones that you have.
Like I have a webcam microphone.
I have this Hyper S microphone Hyper X microphone.
And so I can choose one of those.
And I can just set a voice activation threshold.
Right now I have mine set to 36%.
I'll get to that in a second.
And then I just say load noise torch.
And what happens is I see a new microphone show up in my sound settings that is called noise torch microphone for Hyper X quadcast S analog stereo.
And so I can choose that microphone.
And then that's when it switches over and starts doing the noise suppression.
So it's pretty easy to use.
It will require a little bit of tweaking.
I mean, what you can do is just set your voice activation threshold to, you know, like 80%.
And record a little bit and then load it into audacity or something and kind of see the waveform to see where it's cutting out.
The noise well enough so that it suppresses it.
But also so that it doesn't cut off the beginning of your words.
The higher you set the activation threshold, it will start to cut off the beginning of your words when you start talking.
Maybe it's doing that a little bit for me right now.
But if you go too low, you'll start hearing the background noise start creeping into the recording.
You might even get some clicking.
So you kind of have to play around with that.
Fortunately, it makes it easy to, you know, reload noise torch whenever you change that.
Just keep in mind that when you do reload noise torch that it'll your Linux sound system settings like in your GUI are going to change back to your default microphone.
So you have to remember to switch back to the noise torch microphone.
Otherwise, you're not getting the benefit.
So this can be used in a pulse audio and pipe wire under Linux.
Type wire is a new sound system for Linux.
I'm using pulse audio right now, but I've been thinking about trying a pipe wire soon.
There's a really great YouTube video of a guy also demoing noise torch.
And he runs a hairdryer right, you know, pretty close to the microphone.
He runs a hairdryer while he's talking and you can't hear the hairdryer.
Pretty amazing.
And like I said, you know, my daughter's been playing piano in the background and it's fairly loud for me down here in my basement office.
And otherwise, it would kind of prevent me from doing a podcast at this time.
So yeah, for me, 36, you know, 27%, 36% to 40% works pretty well as a level to set to.
One thing you might notice when you look up noise torch is that there was a security incident a few years ago,
where the guy who maintained noise torch had his computer was compromised.
And so as a precaution, he actually shut down noise torch as a project because he felt like it was the responsible thing to do,
which I think is really great, you know, incident response, you know, care about your users and stuff and he did.
Unfortunately, he didn't really come back to it or he helped out the new project or something.
So some of the community members took the code and they vetted the code.
They went through and did an analysis to look for malicious code and they couldn't find anything.
So it's suspected that, you know, the code is no more dangerous than any other code that you download for your computer.
At this moment, you know, it's all code potentially has some compromise in it, but they couldn't find anything.
So it looks like it's going to be okay generally.
So yeah, check out noise torch and see how it works out for you and hopefully that will enable you to make more podcasts and contribute to HPR more.
Thank you. Have a good day. Bye.
You have been listening to Hacker Public Radio at Hacker Public Radio does work.
Today's show was contributed by a HPR listener like yourself.
If you ever thought of recording podcasts, then click on our contribute link to find out how easy it leads.
Hosting for HPR has been kindly provided by an honesthost.com, the Internet Archive and our sync.net.
On this otherwise stated, today's show is released under Creative Commons, Attribution 4.0 International License.