Prototyping/2019-2020/T3
Prototyping session Spring 2020, in the context of Special Issue #12: Radio Implicancies
This page is currently being worked on.
01
In the work for the Special Issue, we will employ a strategy of working between systems
Proposition is to publish prototyping work alongside other developments of the Special Issue in a rolling fashion, ie following an "early and often" approach to publishing.
Focus on using the wiki as a "cookbook" -- a place to record useful techniques for others.
- Review setup with liquidsoap and icecast from the sandbox to the lurk server.
- Review setup of webhook connecting git repo with webpage on sandbox
- Example of adding a makefile to the workflow. Idea of separating sources from outputs (targets in make speak).
- Connecting to the shared terminal with tmux
So specifically:
ssh sandbox tmux -S /tmp/radio attach -t radio -r
Where "ssh sandbox" assumes you have a "sandbox" entry in your .ssh/config file.
.gitignore
We will follow a "sources only" approach to using the git whereby only "source" files will get committed to a repo. The idea is that the tools are in place for anyone who clones the directory to "build" the other files using the makefile + tools.
For this reason the .gitignore file contains:
*.html
So that if you try to:
git add README.html
You will see:
The following paths are ignored by one of your .gitignore files: README.html Use -f if you really want to add them.
Why .gitignore
- Lighter/smaller repos, faster to share, with less need for huge disk space just to change.
- Less conflicts -- many "artificial" conflicts arise when files that get generated are also version controlled. It's simpler to track only the sources, and let programs like make take care of updating the rest from there.
more .gitignore
See the "live" .gitignore
- Also ignoring media files (use scp to place... manage differently) -- goal: keep the core repository as "light" as possible. Git is not idea for managing big binary files (they don't then to change in a way that lends itself to diff-ing).
- Temporary files (things like foo~)
- Python derived files / cache (python generates __pycache__ folders as needed).
(why) make radio
Why use make? Makefiles are like "executable notebooks" are a really effective (and time tested) way to document and share scripts, workflows, and workarounds using diverse tools.
What's the catch? Make comes from free software programming culture and was created to support programming work rather than publishing. However, as managing programmers is like "herding cats", the tool is extremely flexible and porous, and well integrated with the command-line and BASH scripting. As such it's compatible with the richness of command-line tools available.
- git and make not just for code presentation about using makefile and git in publishing workflows
practical to do's
- Update issue.xpub.nl to point http://issue.xpub.nl/12/ to (redirect to?) The working space of SI12.
- Understand and make use of the git & make based workflow.
- Familiarity using liquidsoap to create icecast streams.
- Use liquidsoap to create flexible streams that can be interrupted
- Use top to check the load on the sandbox
- Use liquidsoap as an editing tool to assemble, mix, and filter audio
- Explore some basic structural elements of liquidsoap (playlist, rotate, sequence).
Make Noise
- the colors of noise (using sox to generate white, brown, and pink noise)
Playlist
Liquidsoap's playlist
Try...
- reload
- reload_mode
Exercise:
1.
ssh to the sandbox
cd cp /var/www/html/12/liquidsoap/playlist.liq . cp /var/www/html/12/liquidsoap/playlist.m3u . cp /var/www/html/12/liquidsoap/passwords.liq .
Copy the whole folder somewhere
cp -r /var/www/html12/liquidsoap .
or
git clone https://git.xpub.nl/XPUB/SI12-prototyping.git
(read-only)
Edit it with nano... Make sure you change the log file to be specific to you...
set("log.file.path","/tmp/<script>_YOURNAME.log")
and try to run it...
liquidsoap -v playlist.liq
2. Clone the respository to your laptop...
apt install liquidsoap
brew install liquidsoap
git clone ssh://git@git.xpub.nl:2501/XPUB/SI12-prototyping.git
Change the playlist to list local audio files... (NB: they need to be stereo)
Change "radioimplicancies.ogg" to whatever name you want...
Live or Fallback
- liquidsoap fallbacks
- mixing diverse sources of material...
- working with silence
- dynamics compression and loudness
- test streaming to an interruptable mount point...
- BASIC cutting / format manipulation with ffmpeg such as extracting chunks with a start / duration
Mixing audio sources
- Filtering for radio/streams:
- compression (as in Dynamic range compression different from audio data compression
- "Loudness" and "Normalisation"... What is replaygain and why can it help?
- liquidsoap for sound processing
Resources
- About the materiality of digital audio: samples, sampling rate, sample width, format, channels... See Sound and listening
- Playing "raw" and non-sound data ... See Sonification
- field recordings @ freemusicarchive
- William Burroughs: cut-ups + The Electronic Revolution
- Example: spectrumsort
- sound art on ubu
Assignment(s) for next week
- Make a "self-referential" radio program with liquidsoap... that is one where the content and the material of its transmission and its tools of production are reflected.
- Document your use of liquidsoap, or other tool(s) in the xpub wiki in the style of a (code) "Cookbook" recipe.
Listen together
- Listen: Alvin Lucier's I am Sitting in a room
- Listen: Janet Cardiff A Large Slow River
- Listen/watch: Laurie Anderson microphone piece
- Listen/watch: Gary Hill
Also for next week, check out Radio Aporee.
02
https://pad.xpub.nl/p/prototyping190520
Set of things to work on / research, make small groups, work on the tasks, report back
- Play/testing an interuptable stream
- How do we archive a liquidsoap program (adding a file as output to the stream)
- Render a waveform with ffmpeg
- "Simple" interface (mapping timeupdate to SVG/canvas)
Exercise (afternoon)
try to type in the following, reading.. Using the "player code"
ssh sandbox cd mkdir -p public_html cd public_html
look at: https://hub.xpub.nl/sandbox/~YOURSANDBOXUSERNAME/
You should see your "homepage" on the sandbox (maybe empty)... Copy the player folder to your public_html...
cp -r /var/www/html/12/player/ . cd player
now you should see the files here:
https://hub.xpub.nl/sandbox/~YOURSANDBOXUSERNAME/player/
ls cat numbers.txt cat numbers.txt | bash -x speaklines.sh ls
cat speaklines.sh
(if you want to modify the numbers)
nano numbers.txt
then...
cat numbers.txt | bash -x speaklines.sh
again...
LOOK at your folder from the browser: https://hub.xpub.nl/sandbox/~YOURSANDBOXUSERNAME/player/
ogginfo speech01.ogg ogginfo speech02.ogg
ls speech*.ogg ls speech*.ogg > counting.m3u cat counting.m3u cat counting.file.liq
liquidsoap -v counting.file.liq
Take a look at counting.ogg (in the browser)
ogginfo counting.ogg
Take a look at dream.html >>>>>>>>>>>> anyone else here??me i opened it in my browser. is that right?yes
Open the console and look at debugging messages while you scroll around... > what is the console? like inspect element console? on firefox it s Tools>Web developer>web console oh yes same thing, different road leading to it. what should we see in console? text while the video is playing, you see the subtitles i neverminf i see it :) woop wooop
(ctrl-shift-I or ctrl-shift-k depending on your browser ctrl-option-j chrome mac) > this is the shortcut didnt work for me tried both?yupbut you're there?yupeh!
ogginfo counting.ogg | python3 ogginfo-to-srt.py --json ogginfo counting.ogg | python3 ogginfo-to-srt.py ogginfo counting.ogg | python3 ogginfo-to-srt.py --vtt
and finally
ogginfo counting.ogg | python3 ogginfo-to-srt.py --vtt > counting.vtt
cp dream.html counting.html
Adjust the html to use counting.ogg and counting.vtt (no path, just the filename) Change the video tag to an audio tag...
links:
https://trac.ffmpeg.org/wiki/Waveform
try
ffmpeg -i counting.ogg -filter_complex "showwavespic=s=1024x120:colors=000000" -frames:v 1 -y counting.png
Notes
- "Live elements" + ways to integrate (metadata) with player page.
- "Radio programs" with liquidsoap... for instance, mixing a set of quotes, music or how might "jingles" function in a longer stream... What structural elements exist in liquidsoap to compose a sequence (like rotate, mix)
- Review / extend the between systems diagram -- Q: How to design a useful system for the metadata.
- Frames (and iframes) ... a brief history? Window.postMessage
- XSS/"Cross-site scripting": Using document.domain, CORS, cgi/php relays, JSONP
- The difficult history of "shared whiteboards" ... Eventual consistency http://www.processwave.org/ recent ipad app VP miro might also look at alternatives like hotglue or YWOT -- projects that are obscure (why? "artistic" nature/lack of entrepreneur-speak, lack of "cutting edge" technology, radical take on collaboration, ultimately trapped in proof of concept, but not really usable?) ... POC with SVG editor and using gitea API as backend ? (is there an API for commmiting files?)... granularity of collaboration ... "realtime" vs. "shared"... what if one at a time editing is OK, given a workable/fluid social protocol around it?
- authentication
- Visualization "sound cloud" style of a rendered stream ... showing structure...
03
RADIO PROTOCOLS aka METADATA, or STREAMING TEXT
https://pad.xpub.nl/p/prototyping260520
Radio Protocols
In a one sense ("media"), radio protocols could be considered the conventions of a radio station, involving structural elements of "programming" in the sense of what gets played when. It involves "formats" (weather report, traffic report, requests), schedules (news bulletin on the hour), and programs (Jazz brunch).
In a informatics/computer science context, protocols are also social agreements made around technical formats (like HTML, ogg/mp3) often involved with how different programs / processes can communicate with each other, often in the context of a network. For instance HTTP as the protocol that occurs between web browser and web server to request and then transmit the contents of a web page. Streaming media has specialized protocols (like ...), though the Icecast server we use uses a variation of HTTP. Icecast also has related protocols called ICY related to managing/requesting the metadata of a stream.
Protocols can range from "ad hoc" best practices / hacks, to highly formalized processes.
Many protocols are based on simple exchanges of text, typically formatted in lines (ie each line is a significant "unit" for transmitting and requesting information from a server.
Based on the question, "how might we share the terminal as part of the broadcast?"... We are going to be looking at how the textual output of the liquidsoap server (1) be transmitted to listeners, and (2) could be filtered and used as a novel protocol for the Radio Implicancies transmissions.
Consider searching for "syncing metadata with icecast"...
preparing and testing a stream with replay gain
How might me make a test stream to hear smaller samples of each element in a playlist?
"Random Access"
Dynamic playlist(s)
- Add (random) excerpts of "raw" materials ?!
- How to combine one or more playlists, with randomness.
- How to use jingles to "join" elements
- How might "bumpers" function to mark beginnings and ends of "programs".
- How might we make the program more inviting for outside/new listeners? (Meta programming?)
regular expressions
pipeserver
Use cue_cut and annotation to prepare a test stream.
- How might we filter the messages that can sent to the stream (how to filter with commandline)
- Add metadata to the audio files
- Add a metadata "reporter" to the liq
- Run liq + pipe to pipeserver
- Adjust pipeserver.html
commandline calisthenics
ls *.ogg | wc -l ls *.ogg > playlist.m3u cat playlist.m3u | wc -l
# multitasking commandline style
nano foo.txt ctrl-z fg
ctrl-z ps
# PID are process ids Note which one is nano then...
fg XXX
ps aux | grep liquidsoap
creating a custom pipeserver interface
cd cd public_html cp /home/mmurtaugh/pipeserver/pipeserver.html .
nano pipeserver.html
Add
<audio controls src="http://echo.lurk.org:999/radioimplicancies_test.ogg"></audio>
Change line 31 to look like:
var ws_addr = 'ws://'+window.location.host+'/pipe',
notes
- replaygain, Loudness normalization vs. Dynamic normalisation / compression: script to test first x sec each source... just liquidsoap?
- metadata hook
- liquidsoap server? / telnet?
- countdown ... (say protocol, add?)
- vorbisgain is a tool following the ReplayGain algorithm, that adds metadata to an OGG audio file describing the loudness of the file. It was created for music players to be able to keep a consistent audio level between different albums.
- pipebot?
- materiality of timed text ... epicpedia ... etherlamp...
- chat archive to vtt
- podcast feed ?
broadcast notes
Set a log file on radio2:
set("log.file",true) set("log.file.path","<script>.log") set("log.stdout",true)
Follow the log with tail -f (follow) and pipe to the pipeserver:
tail -f 12.2.livetest.log | pipeserver.py --host 0.0.0.0
shell.html
aka pipeserver frontent...
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8">
<title>shell</title>
<style>
body {
background: #888;
}
body.connected {
background: white;
}
#shell {
white-space: pre-wrap;
height: 25em;
width: 100%;
font-family: monospace;
color: white;
font-size: 14px;
background: black;
overflow-y: auto;
}
</style>
</head>
<body>
<p><span id="connections">0</span> active connections</p>
<div id="shell"></div>
<script>
var ws_addr = 'wss://'+window.location.host+"/12/pipe/",
sock = null,
shell = document.getElementById("shell"),
connections = document.getElementById("connections");
function connect () {
sock = new WebSocket(ws_addr);
sock.onopen = function (event) {
console.log("socket opened");
document.body.classList.add("connected");
// sock.send(JSON.stringify({
// src: "connect",
// }));
};
sock.onmessage = function (event) {
// console.log("message", event);
if (typeof(event.data) == "string") {
var msg = JSON.parse(event.data);
// console.log("message JSON", msg);
if (msg.src == "stdin" && msg.line) {
// Show all lines
var line = document.createElement("div");
line.classList.add("line");
line.innerHTML = msg.line;
shell.appendChild(line);
// scroll to bottom
shell.scrollTop = shell.scrollHeight;
} else if (msg.src == "connect") {
connections.innerHTML = msg.connections;
}
}
};
sock.onclose = function (event) {
// console.log("socket closed");
connections.innerHTML = "?";
document.body.classList.remove("connected");
sock = null;
window.setTimeout(connect, 2500);
}
}
connect();
</script>
</body>
</html>
live_interrupt_withmetadata.liq
"RADIO1" program, with a callback to recordmeta.py, a script that relays the data (via POSTign to a cgi script on (issue.)xpub.nl)
# /var/www/radio/live_interrupt_withmetadata.liq
set("log.file",false)
# set("log.file.path","/tmp/<script>.log")
set("log.stdout",true)
%include "/srv/radio/passwords.liq"
# Add the ability to relay live shows
radio =
fallback(track_sensitive=false,
[input.http("http://echo.lurk.org:999/radioimplicancies_live.ogg"),
single("brownnoise.mp3")])
###################
# record metadata
scripts="/var/www/html/radio/"
def on_meta (meta)
# call recordmeta.py with json_of(meta)
data = json_of(compact=true,meta)
system (scripts^"recordmeta.py --timestamp --post https://issue.xpub.nl/cgi-bin/radioimplicancies.cgi "^quote(data))
# PRINT (debugging)
#print("---STARTMETA---")
#list.iter(fun (i) -> print("meta:"^fst(i)^": "^snd(i)), meta)
# data = string.concat(separator="\n",list.map(fun (i) -> fst(i)^":"^snd(i), meta))
#print ("---ENDMETA---")
end
radio = on_metadata(on_meta, radio)
########################################
radio = mksafe(radio)
output.icecast(%vorbis,
host = ICECAST_SERVER_HOST, port = ICECAST_SERVER_PORT,
password = ICECAST_SERVER_PASSWORD, mount = "radioimplicancies.ogg",
radio)
recordmeta.py
#!/usr/bin/env python3
import sys, argparse, datetime, json
import requests
# todo: add timestamp
ap = argparse.ArgumentParser("")
ap.add_argument("data")
ap.add_argument("--post")
ap.add_argument("--logfile")
ap.add_argument("--timestamp", default=False, action="store_true")
args = ap.parse_args()
data = json.loads(args.data)
if args.timestamp:
data['time'] = datetime.datetime.utcnow().isoformat()
if args.logfile:
with open (args.logfile, "a") as f:
print (json.dumps(data), file=f)
if args.post:
r = requests.post(args.post, json=data)
resp = r.content.decode("utf-8").strip()
print ("POST response ({}): {}".format(r.status_code, resp))
radioimplicancies.cgi
liveplayer.js
function liveplayer (elt) {
var last_timestamp,
DELAY_TIME_SECS = 5;
async function poll () {
var rows = await (await fetch("/cgi-bin/radioimplicancies.cgi")).json();
// console.log("rows", rows);
if (rows) {
// get the last (most recent) item
var d = rows[rows.length-1];
if (d.time && d.time != last_timestamp) {
last_timestamp = d.time;
console.log("liveplayer: current metadata", d);
window.setTimeout(function () {
var old_nowplaying = elt.querySelector(".now_playing");
if (old_nowplaying) {
old_nowplaying.classList.remove("now_playing");
}
var nowplaying = elt.querySelector("#"+d.title);
if (nowplaying) {
console.log("liveplayer: nowplaying", nowplaying);
nowplaying.classList.add("now_playing");
} else {
console.log("liveplayer: warning no div matching title " + d.title);
}
}, DELAY_TIME_SECS*1000);
}
}
}
window.setInterval(poll, 1000);
}
04
xtended PLAYERS / TIME BASED TEXT(s) / REPLAYING (textual) stream
FROM last time, form groups around the following...
- ARCHIVE of the liquidsoap log: How to replay this text in sync with the archive?
- (Regenerate program from the playlist => OGG + LOG)
- logplayer.js to play the log synced with audio recording.
- ARCHIVE of the etherpad https://pad.xpub.nl/p/radio_implicancies_12.3/export/etherpad
- playing the etherpad
- Image representation of stream + javascript (archive visualisation) / player
- waveform + alternatives ... what other kinds of time based traces make sense... (
- (live) Audio visualization (HTML5 audio + canvas) (proxy the audio via xpub / nginx ?)
... =
LIVENESS
- What possibilities exist in LIVE situations... (alternative representations of presence / writing systems / CHAT?)
- Image representation of stream + javascript (archive visualisation)
- Live audio visualization (HTML5 audio + canvas) (proxy the audio via xpub / nginx ?)
PLAYER
Radio webpage
- Audio player (html/js)
- META DATA display / COUNTDOWN
- INTEGRATION with chat / etherpad ?? ...
- iframes
- guest book cgi??!! (socially determined captcha !)
- Player: display metadata (Metadata from icecast? https://stackoverflow.com/questions/6115531/is-it-possible-to-get-icecast-metadata-from-html5-audio-element ) TRY!
- Generate podcast/RSS with json + Template (template.py(python+jinja)/makefile)
- Stream access (scheduling)
(meta/way of working Publishing via git/githook TRY! )
- random access media ?? (can an mp3 do a range request jump? or only ogg ??) depends on server, example of non-random access mp3
- ubu audio player as anti-pattern ?!
- html5 media fragments, and javascript
- synced visualisations ???... html triggered by stream / metadata... (or not ?!)
VOICE
- text to speech / speech synthesis
Specifically the case of [MBROLA](https://github.com/numediart/MBROLA) and the "free" but not "libre" voices. https://github.com/numediart/MBROLA-voices
Espeak + MBrola https://github.com/espeak-ng/espeak-ng/blob/master/docs/mbrola.md
Projects spɛl ænd spik. hellocatfood (antonio roberts) https://www.hellocatfood.com/spell-and-speak/
"real" VOICES http://librivox.org
- speech to text
- We Hope This Email Finds You Well
- festival provides *text2wave* which liquidsoap uses
- microphones
HRTF
- ASMR - HRTF ... https://www.youtube.com/watch?v=3VOLkehNz-8
- Spatilsation of Audio: https://www.york.ac.uk/sadie-project/database.html, head-related transfer functions (HRTFs)
- CLI spatialisation of audio: ffmpeg headphone, sofalizer
- https://trac.ffmpeg.org/wiki/AudioChannelManipulation#VirtualBinauralAcoustics
BOTS
- "posthuman" ... hayles
Cross system / bots (2 parts?)
- infobot
- eliza
- mud/moo
- Bot / chatroom
- Radio stream / schedule
- mediawiki API (pull + push) .. example of epicpedia?
- etherpad API (pull + push?) .. example of etherlamp (make js version?)
- Example of live log of chat , replayed with video recording...
See: https://vnsmatrix.net/projects/corpusfantastica-moo-and-lambda-projects
LEAVE A MESSAGE / RECORDER
See Leave a message
- https://mdn.github.io/web-dictaphone/ https://github.com/mdn/web-dictaphone/ (+ javascript looping / instrument) ... now how to save / submit ... (
- online recording interface... interface with CGI to save/serve files ?
- https://developer.mozilla.org/en-US/docs/Web/API/MediaRecorder
- https://github.com/mattdiamond/Recorderjs
- Angeliki Diakrousis http://radioactive.w-i-t-m.net/
- IDEA: Limit recordings to a fixed duration (data size)...
UNCATEGORIZED ;)
- gitea webhooks
- CGI to receive a JSON post
- php to receive JSON post (example on gitea webhooks page)
- CGI (making a "helloworld" / printenv with python) See CGI checklist
- Don't forget "sudo a2enmod cgi" and "sudo systemctl restart apache2" to enable the cgi module in Apache...
MIDI
- David Cope https://computerhistory.org/blog/algorithmic-music-david-cope-and-emi/ http://artsites.ucsc.edu/faculty/cope/5000.html
- Corneluis Cardew: Towards an Ethic of Improvisation The Great Learning
SUPERCUTS
- Example of marking up moments of stream / media links (in etherpad ... or outside player ??)... POC: js to make media links...
- Complementary ffmpeg commands to extract said portions...
- EXAMPLES of making supercuts of material...
- TEMPORAL TEXT: VTT
- TEMPORAL TEXT: etherpad ... js player of "raw" etherpad...
- EXERCISE: make COUNTDOWNS (with liquidsoap ?? + espeak)
- work with schedules? timing/?
- AI
- ethercalc to organize things ... examples of export to JSON / CSV / HTML through template
- NORMALIZATION test ... try some examples...
- XMPP + links to images ?!
- listener relationships...
- TV! (picture ?! / visuals ??) how to broaden beyond voice... VISUAL component...
- LIVE GUESTS / "call in" / voice mail, prompts,
- Sockets / FIFO style pipeline bots ??
- archiving of the previous week...
- listening to data(sets)
- test say_metadata