Whisper Radio

IceCast radio station of a TTS (espeak) reading off:

• Random text file
• RSS headline
• Weather data (METAR)
• Random music file
• Reads the "freshest" thread on gopherden (gopher://gopher.someodd.zip/phorum)
• MOTD: a message you want to consistently be read
• Some Masstodon/Fosstodon data
• Plays Drum & Bass (90s to early 2000s) from Internet Archive
• Responds to the latest post under a specific hashtag on fosstodon (uses AI)

"Whisper" is a reference to the espeak voice or whatever used.

Please also see the showcase post I made about this on my blog. It has some tips for troubleshooting and setting up an Icecast2 server.

Here's a demo/the official Whisper Radio stream, you may be able to open in your browser or a media player: https://radio.someodd.zip/stream

Be sure to do this then edit the files before you begin:

cp ezstream.example.xml ezstream.xml
cp config.example.sh config.sh

Now being used by Bitreich!

Warning

This was an experiment in seeing how far I could take shell scripting, kinda, and it's just a real headache sometimes. There may be bugs!

Dependencies

• espeak
• jq
• xmlstarlet
• curl
• ezstream
• icecast2
• metar
• piper-tts
• just take a look at ctts.py if you wanna use it (used by default)
• ffmpeg

sudo apt-get update
sudo apt-get install espeak jq curl ezstream icecast2 metar ffmpeg pipx ffmpeg xmlstarlet
pipx install piper-tts

Content setup

Make sure to create and populate these directories in the project folder:

• audio/ put mp3s and the like in here. The pool of audio that'll be randomly selected from to build the radio program.
• text/ same as above, but with plain text files.

Also check config.sh for Atom/RSS feeds, gopher page, etc.

piper-tts setup

Download:

• The JSON for the hfc_female Piper voice
• The hfc_female Piper voice
• Check out the supported voices for Piper

You may have to open a new terminal for it to be in the path.

As of right now I had to use this weird setup to get around a current issue with piper installation:

pipx install uv
uv tool install --python 3.11 piper-tts==1.2.0

AI TTS (Cloned Voice) Setup

Just read ctts.py.

icecast2 set up

Install icecast2. It asked me to optionally input some info through the TUI. For my example/demo I chose hackme (source password, relay password, admin password [this is extremely bad practice]). I also entered localhost.

sudo apt-get update
sudo apt-get install icecast2

Edit sudo vi /etc/icecast2/icecast.xml.

sudo systemctl start icecast2

To persist the server:

sudo systemctl enable icecast2

I RECOMMEND YOU USE LOW LOW LOW MP3 STREAMING SETTINGS, MONO ETC

Copy the repo

git clone https://github.com/someodd/whisper-radio
cd whisper-radio

Run the script for first time + send with ezstream

./whisper.sh

Crontab

Use crontab -e, run every hour and log errors:

0 * * * * /home/baudrillard/Projects/whisper-radio/whisper.sh 2>> /home/baudrillard/Projects/whisper-radio/logfile

Troubleshooting

I had this problem where I thought the project was broken but I restarted computer and it works again. careful how kill ezstream maybe.

Tips

Link music from an archive of some musician to the possible audio sources, to avoid duplication:

sudo find /foo/bar/some-legal-music-archive -name "*.mp3" -exec ln -s {} ./audio/ \;