Night Vale Numbers Station
A streaming radio server playing a Night Vale-esque numbers station.
Serving Audio
Launch the server via:
docker-compose up
Icecast server will be streaming on host port 8000 with a single stream at /nightvale.mp3.
Ezstream source client will dynamically generate an mp3 playlist and stream it to the Icecast server.
The Icecast admin UI is available at http://localhost:8000. Admin credentials are admin/nightvale.
Listening
Web Client
A small webserver is deployed automatically with the audio server, and is serving from http://localhost:8080. It's serving the contents of the ./client directory, which includes a simple HTML-based web client pointing at the local stream.
Other Clients
Connect to Icecast from a local client (e.g. VLC) at http://localhost:8000/nightvale.mp3 to hear the stream.
Development Notes
Text-to-speech
I tried two different ways to generate the mp3 files of spoken numbers in the ./data directory.
generate_numbers_docker.sh uses espeak to generate audio and sox to encode to mp3, performed in a docker container.
generate_numbers_mac.sh uses OSX's say to generate the audio and lame to encode to mp3, performed on the host. Requires lame package from homebrew.
See comments in the scripts for more info.
Playlist generation
The ezstream source client for Icecast is responsible for generating an mp3 stream and sending it to the Icecast server.
I'm using the Playlist Program feature of ezstream to generate a pseudorandom mix of the source mp3 files, via the ./ezstream/next_playlist_item.sh script. See comments in the script for more info.
Audio Files
Helpful tools:
afplay to play an audio file from the command line. (OSX)
afinfo to inspect an audio file and determinte bitrate, sample rate, channels, etc. (OSX)
Chime sound
I found it on the internet. Had to change the channel count and sample rate to match the other mp3 files or the stream clients would break/stop playing when the chime appeared in the stream.
~/src/me/nightvale/data $ docker run -it --rm -v $(pwd):/data ozzyjohnson/tts /bin/bash
root@45bde6699aa3:/data# sox chimeraw.mp3 -r 16000 -c 1 chime.mp3
Also had to remove ID3 tags from the chime file:
brew install eyed3
eyeD3 --remove-all chime.mp3
TODO
- Add high-cut and low-cut filters to make the audio sound like an AM radio station
- Add background static
- Add a very rare playlist item where the announcer breaks character to give an ominous warning or emotional breakdown, and is cut off abruptly, followed by extended silence
- Find a better chime