Alright, I wrote a blogpost about my adventures:

This doesn't have much new compared to my posts from this weekend, but here they are ordered and coherent.

I also posted some audio samples:

Keep in mind that both samples will sound pretty much horrific, but also that each sample is around 100kb for more than a minute of audio.

I prefer the Opus file to the mp3 file, as it's a little less muddy during speech (but ruins music.)

@Ethancdavenport - I feel like this might actually interest you so I wanted to make sure to show it to you.

Something to consider re: - Average dialup bitrate (towards the end of the dialup era) was 40–50 kbit/s, right?

These files are averaging about 9kbps.

You could stream a on a dialup connection without saturating the connection.

If that sounds like a ridiculous thing to care about, consider that the easiest way to increase the distance over which you can transmit a signal is to decrease the transmission speed.

So, if you were hypothetically trying to build a new network that transferred media over long range radio connections, phone lines, and sneakernets, very small file sizes would be very beneficial.

@ajroach42 Another way to look at it: 9 kbps is slow enough to transfer over 9600 bps AX.25 packet radio, in *real time* (although you're not using stuff ideally suited for degraded-link comms, so packet loss would be nasty, especially given the notorious unreliability of 9600 AX.25).

@ajroach42 Not enough thought gets put into the fact that we'll someday be wanting to have conversations with Mars.

@emsenn @ajroach42 There was a Sci Fi story on that score. To improve communications, talk like women can. Both sides talk continuously and answer questions as they arrive as well. It removes the limitation of waiting a few minutes delay to answer a question by keeping the information flowing continuously.


Yeah, isn't that how FT8 call works? That can work below the noise floor by averaging across a repeated signal.

@ajroach42 I really liked your floppycast post. I do pico-podcasts which are deliberately under 2 minutes. Still trying to get the bit rate down though 😊

@Edent Neat!

I'll be happy to offer some tips that don't absolutely ruin your sound quality if you'd like.

@ajroach42 I'm assuming that iPhone won't play back opus files? Or have they gained that ability since I last had one?

@Edent I don't have any first hand experience, as I don't have an iPhone anymore, but according to the internet: "Opus is now included in iOS 11."

@ajroach42 Sweet! Thanks! Also, can I interview you for my podcast? Will literally take less than 5 minutes 😊

Maybe! When, and via what platform?

Where can I hear your podcast?

@ajroach42 podcast is at
You can either send me a 60 second audio clip, or we can record via Skype.

@Edent Either is fine with me. I use Zencaster, if that would be easier.

I'll have to find time, though. 6-7 hours of meetings today. :-D

@ajroach42 I'm abroad for the next few days! I'm sure we can figure out a mutually inconvenient timezone 😂

@ajroach42 POTS audio is only 64Kbps, uncompressed. I'm surprised modern compression can't get way better than that.

@tek Oh, we can. If you want to devote some serious power to playback, stuff like Codec 2 can apparently deliver pretty stunning results.

And VoIP codecs can do some pretty interesting stuff, too, but none of those play back on vintage hardware.

@ajroach42 @tek TBH I wonder how much power is necessary - AFAIK they're doing realtime encode and decode on a 160ish MHz Cortex-M4 with time to spare, and that's in a RAM-tight environment.

In a RAM-heavy environment like an old personal computer, you might be able to implement some of the algorithms as lookup tables (if they aren't already) and make it even faster?

@tek Thankfully, it's one I'm free to ignore now that I know what the bottom of the barrel sounds like for MP3 and Opus.

My next attempt will be codec2.

@ajroach42 feels kinda weird that I missed this post when you posted it, and only knew to go back and look for it because of the HN link. :awesome:

Just read your blog post on #floppycasts - it sounds like a really interesting project. Personally I prefer the idea of a shorter #floppycast with better audio quality.

I've been podcasting for over 11 years now, and I keep threatening to do a talk #shotcast - so I might see if I can combine it with this concept to keep the length down.

@thelovebug I definitely looked in to that as an option as well.

Combined with some of the other techniques one would normally use to decrease size without ruining quality, it's possible to get decent sounding audio of up to 10 minutes in length on a floppy disk.

When I first looked in to this, my plan was to scrip five minutes, record, and then encode to fit. That way you encourage concision without feeling trapped by two minutes of available recording time.

But I talk a lot.


Would be really nice if #floppycast ing would become a thing, preferably with codec2 and wavenet decoding and fediverse integration.

@jsilence I'm looking in to codec2 this week. From what I've gathered, there's no software that would play it back on my target machines (486 and Pentium era computers running DOS), but beyond that it seems like the perfect fit for this project (and for encouraging more broad adaption of the concepts.)


Contemplating on starting a blogcast about climate change. German media is really neglecting the topic and I'd like to translate, contextualize and comment on english spoken media coverage. Doing this in a #floppycast format would really fit well with the topic of low ressource usage. #formfollowsfunction


Distribution of the #floppycast via dat:// or ipfs would make sense as well.

@jsilence I like the way you think!

I'm still not 100% sold on dat because the only implementation is node.js, but distributing the thing peer-to-peere seems like a good call.

Maybe even over SSB?

Focusing on very small file size has a huge impact on penetration when you're distributing peer-to-peer.

(I really wish there was a podcast client or an RSS reader with torrent based distribution built in. I'd already have a (fairly low bitrate) video podcast in distribution.)


Too little practical experience with dat, ipfs or SSB. Read a piece on HN lately on how ipfs is not yet up to snuff and not well managed. A commenter mentioned that dat is way ahead with that respect.
Propably a #floppycast could be distributed via all those fancy new p2p channels.

@jsilence Dat is the furthest along, and I use it pretty regularly, I just don't like the bloat of a JS application being used in this respect.

I also have some issues with SSB, on several fronts, but it seems like a great avenue for distributing tiny files.

To me, though, it seems like this is a perfect usecase for torrents + atom/RSS feeds.


Ok, will look into SSB. So far I only have read the name in context with p2p a couple of times. Need to investigate. Will be back when I watched all SSB videos on YT. ALL OF THEM!


I see, the #floppycast file would end up in the blobs folders of SSB users, so squeezing file size would be really nice to them.

@jsilence I love that idea!

I don't speak German, but @Ethancdavenport does a bit, maybe he wants to help?

floppydisks are a powerful symbol of progress, waste, recycling, and resource maximization. Even though it's unwieldy and impractical to try to condense down to that size, it also send a message.

@ajroach42 @jsilence I’m rusty as all hell but that sounds like an awesome project!


Codec2 + WaveNet decoding... I am not sure whether I totally get how a production pipeline would look like. Wav to C2 via command line encoder, ok. But then how to train and deploy the decoder component? Would one have to deploy the decoder together with the C2 file? Wouldn't that defeat the purpose of having really small file sizes?


@jsilence @Ethancdavenport I have yet to find the time to research exactly how this would work, so I dunno.

Sign in to participate in the conversation
R E T R O  S O C I A L

A social network for the 19A0s.