A few things

twitter: @JosephErnest
email: here

Articles about:
#music
#photo
#programming

Don't read #tech articles except you really want to.

Some of my projects:
BigPicture
Jeux d'orgues
SamplerBox
Ojourdui

An attempt to generate true entropy and random data with audio (and your computer's built-in microphone)

You probably know that generating some real random data is not so easy to do with a computer. How to design a good Random Number Generator (or a pseudo-random one) is a math topic that you can work years on ; it's also something very important for real-life applications such as security/cryptography, for example when you need to generate strong passwords.

Usually (and this is true in general in cryptography), designing your own algorithm is bad, because unless you're a professional in this subject and your algorithm has been approved by peers, you're guaranteed to have flaws in it, that could be exploited.

But here, for fun (don't use it for critical applications!), let's try to generate 100 MB of true random data.

1) Record 20 minutes of audio in 96khz 16bit mono with your computer's built-in microphone. Try to set the mic input level so that the average volume is neither 0 dB (saturation) nor -60 dB (too quiet). Something around -10 dB looks good. What kind of audio should you record? Nothing special, just the noise in your room is ok. You will get around 20*60*96000*2 ~ 220 MB of data. In these 220 MB, only the half will be really useful (because many values in the signal - an array of 16-bit integers - won't use the full 16-bit amplitude: many integers "encoding" the signal might be for example of absolute value < 1024, i.e. will provide only 10 bits)

2) Now let's shuffle these millions of bits of data with some Python code:

from scipy.io import wavfile
import numpy as np
import functools

sr, x = wavfile.read('sound.wav')  # read a mono audio file, recorded with your computer's built-in microphone

#### GET A LIST OF ALL THE BITS
L = []  # list of bits
for i in range(len(x)):
    bits = format(abs(x[i]), "b")  # get binary representation of the data
                                   # don't use "016b" format because it would create a bias: small integers (those not using
                                   # the full bit 16-bit amplitude) would have many leading 0s!
    L += map(int, bits)[1:]        # discard the first bit, which is always 1!

print L.count(1)
print L.count(0)  # check if it's equidistributed in 0s and 1s

n = 2 ** int(np.log2(len(L)))
L = L[:n]  # crop the array of bits so that the length is a power of 2; well the only requirement is that len(L) is coprime with p (see below)

### RECREATE A NEW BINARY FILE WITH ALL THESE BITS (SHUFFLED)
# The trick is: don't use **consecutive bits**, as it would recreate something close to the input audio data. 
# Let's take one bit every 96263 bits instead! Why 96263? Because it's a prime number, then we are guaranteed that
# 0 * 96263 mod n, 1 * 96263 mod n, 2 * 96263 mod n, ..., (n-1) * 96263 mod n will cover [0, 1, ..., n-1].  (**)
# This is true since 96263 is coprime with n. In math language: 96253 is a "generator" of (Z/nZ, +).

p = 96263  # The higher this prime number, the better the shuffling of the bits! 
           # If you have at least one minute of audio, you have at least 45 millions of useful bits already, 
           # so you could take p = 41716139 (just a random prime number I like around 40M)

M = set()
with open('truerandom', 'wb') as f:
    for i in range(0, n, 8):
        M.update(set([(k * p) % n for k in range(i, i+8)]))  # this is optional, here just to prove that our math claim (**) is true
        c = [L[(k * p) % n] for k in range(i, i+8)]   # take 8 bits, in shuffled order
        char = chr(functools.reduce(lambda a, b: a * 2 + b, c))  # create a char with it
        f.write(char)

print M  == set(range(n))  # True, this shows that the assertion (**) before is true. Math rulez!

Done, your truerandom file should be truly random data!

Notes:

How to create symbolic links with Windows with a GUI (no command-line)?

Quick tip: here is how to create symlinks in Windows without using the command line tool mklink.

Create symbolic links with a GUI

1) If you have Python installed, create mklinkgui.py:

import win32clipboard    # pip install pywin32 if you haven't installed it already
import sys, os, subprocess
fname = sys.argv[1]
win32clipboard.OpenClipboard()
filenames = win32clipboard.GetClipboardData(win32clipboard.CF_HDROP)
win32clipboard.CloseClipboard()
for filename in filenames:
    base = os.path.basename(filename)
    link = os.path.join(fname, base)
    subprocess.Popen('mklink %s "%s" "%s"' % ('/d' if os.path.isdir(filename) else '', link, filename), shell=True)

2) Open regedit and

How to use it?

How to find seamless loops in audio files (with a little bit of math and programming)

When making instrument sample sets (e.g. church organ sample sets used with Hauptwerk or GrandOrgue, see my project Jeux d'orgues), we need to set looping points in WAV audio files:

such that when playing the part [a, b] in loop, we don't hear any click or pop when the sample reaches the end of the loop.

Example 1: bad loop with audible clicks

Example 2: seamless loop with no click, that's what we are looking for! The loop has a ~ 2.670 second period, can you hear where are the looping points?

 

Finding looping points can be done manually but this is a very long and tedious task. A few programs exist to do this process automatically such as Extreme Sample Converter (it has an excellent auto-looping algorithm), LoopAuditioneer (open source), Zero-X Seamless Looper, SampleLooper, etc.

Here we'll look at a home-cooked algorithm that works well to detect looping points.

First of all, let's load the audio file (downloadable here) with Python:

from scipy.io import wavfile
import numpy as np
import itertools

sr, x = wavfile.read('060.wav')
x0 = x if x.ndim == 1 else x[:, 0]     # let's keep only 1 channel for simplicity, but we could easily generalize this for 2 channels
x0 = np.asarray(x0, dtype=np.float32)

Let's say the audio file's sustain part (this is precisely where we're looking for a loop!) begins at t=2 sec and finishes at t=9 sec. We will now subdivide the time-interval [2 sec, 9 sec] into a 250 milliseconds grid: 2, 2.25, 2.5, 2.75, 3, 3.25, ..., 8.75, 9.

From this sequence, we now create "loop candidates" (a, b) of length at least 1 second, example: (2.5, 7.5), (3.25, 5.75), (6.0, 8.75), etc. Then, for each loop candidate, we'll improve the loop (this is the core of the algorithm, it will be discussed in the next paragraph) and compute a distance d. We finally keep the loop that has the minimal distance (among all loop candidates). Finished!

A = [int((2 + 0.25 * k) * sr) for k in range(29)]  # the grid 2, 2.25, 2.5, ... 8.75, 9
dist = np.inf
for a, b in itertools.product(A, A):  # cartesian product: pairs (a, b) of points on the grid
    if b - a < 1 * sr:
        continue
    a, B, d = improveloop(x0, a, b, sr=sr)
    print 'Loop (%.3fs, %.3fs) improved to (%.3fs, %.3fs), distance: %i' % (a * 1.0 / sr, b * 1.0 / sr, a * 1.0 / sr, B * 1.0 / sr, d)

    if d < dist:
        aa = a
        BB = B
        dist = d 

print "The final loop is (%.3fs, %.3fs), i.e. (%i, %i)." % (aa * 1.0 / sr, BB * 1.0 / sr, aa, BB)

Finished? Not yet! We need to explain what we mean by improving a loop, as that's the crucial part of the algorithm. More precisely, we'll now explain how to transform a loop (3.25, 5.75) with points taken on the grid (this random loop probably "clicks" like in Example 1 before!) into a "good loop" (3.25, 5.831). Let's zoom on the junction point to understand what's going on:

How to measure if a loop is good or not? Ideally, if the loop (a, b) is perfect/seamless, x[a:a+10 ms] should be very close to x[b:b+10 ms]. Measuring how close two arrays x and y are can be done by computing sum((x[n]-y[n])^2), and if the sum is small, x and y are close.

Finding k such that np.sum(np.abs(x0[a:a+W1]-x0[k+b:k+b+W1])**2) is minimal can be obtained by noting that

(x[n] - y[n+k])**2  = x[n]**2 - 2*x[n]*y[n+k] + y[n+k]**2

and by using numpy.correlate. We can now define this function:

def improveloop(x0, a, b, sr=44100, w1=0.010, w2=0.100):
    """
    Input:  (a, b) is a loop
    Output: (a, B) is a better loop 
            distance (the less the distance the better the loop)
    This function moves the loop's endpoint b to B (up to 100 ms further) such that (a, B) is a "better" loop, i.e. sum((x0[a:a+10ms] - x0[B:B+10ms])^2) is minimal
    """

    W1 = int(w1*sr)
    W2 = int(w2*sr)
    x = x0[a:a+W1]
    y = x0[b:b+W2]
    delta = np.sum(x**2) - 2*np.correlate(y, x) + np.correlate(y**2, np.ones_like(x))
    K = np.argmin(delta)
    B = K + b
    distance = delta[K]

    return a, B, distance

That's it, in less than 50 lines of Python code!

This audio file

(looped 4 times here but we could loop it forever) has been obtained with the algorithm described here. Not too bad, n'est-ce pas?


Example of output:

Loop (2.000s, 3.000s) improved to (2.000s, 3.009s), distance: 1003724800
Loop (2.000s, 3.250s) improved to (2.000s, 3.340s), distance: 839278592
Loop (2.000s, 3.500s) improved to (2.000s, 3.559s), distance: 1281863680
[...]
Loop (2.000s, 8.500s) improved to (2.000s, 8.544s), distance: 1092337664
Loop (2.000s, 8.750s) improved to (2.000s, 8.789s), distance: 964747264
Loop (2.000s, 9.000s) improved to (2.000s, 9.004s), distance: 2488913920
[...]
Loop (7.750s, 9.000s) improved to (7.750s, 9.004s), distance: 1167093760
Loop (8.000s, 9.000s) improved to (8.000s, 9.001s), distance: 1710333952

The final loop is (6.750s, 8.322s), i.e. (297675, 366989).

Note: Wouldn't it be possible to save these loop markers inside the WAV file's metadata instead of just printing them on screen? Sure it is, but as Python's standard library doesn't support WAV markers editing, you'll have to use these techniques to do this.

Working with audio files in Python, advanced use cases (24-bit, 32-bit, cue and loop markers, etc.)

Python comes with the built-in wave module and for most use cases, it's enough to read and write .wav audio files.

But in some cases, you need to be able to work with 24 or 32-bit audio files, to read cue markers, loop markers or other metadata (required for example when designing a sampler software). As I needed this for various projects such as SamplerBox, here are some contributions I made:

  1. The Python standard library's wave module doesn't read cue markers and doesn't support 24-bit files. Here is an updated module:

    wave.py (enhanced)

    that adds some little useful things. (See Revision #1 to see diff with the original stdlib code).

    Usage example:

    from wave import open
    
    f = open('Take1.wav')
    print(f.getmarkers())

    If you're familiar with main Python repositery contributions (I'm not), feel free to include these additions there.

  2. The module scipy.io.wavfile is very useful too. So here is an enhanced version:

    wavfile.py (enhanced)

    Among other things, it adds 24-bit and 32-bit IEEE support, cue marker & cue marker labels support, pitch metadata, etc.

    Usage example:

    from wavfile import read, write
    
    (sr, samples, br, cue, cuelabels, cuelist, loops, f0) = read('Take1.wav', readmarkers=True, readmarkerlabels=True, readmarkerslist=True, readpitch=True, readloops=True)
    print read('Take1.wav', readmarkers=True, readmarkerlabels=True, readmarkerslist=True, readpitch=True, readloops=True)
    
    write('Take2.wav', sr, samples, bitrate=br, markers=cue, loops=loops, pitch=130.82)
    print read('Take2.wav', readmarkers=True, readmarkerlabels=True, readmarkerslist=True, readpitch=True, readloops=True)
    
    write('Take3.wav', sr, samples, bitrate=br, markers=cuelist, loops=loops, pitch=130.82)

    Here is a Github post and pull-request about a hypothetical merge to Scipy.

Here is how loop markers look like in the good old (non open-source but soooo useful) SoundForge:


Lastly, this is how to convert a WAV to MP3 with pydub, for future reference. As usual, do pip install pydub and make sure ffmpeg is in the system path. Then:

from pydub import AudioSegment
song = AudioSegment.from_wav("test.wav")
song.export("test.mp3", format="mp3", bitrate="256k")

will convert a WAV file to MP3.

Make a zooming + panning user interface work on mobile devices (in progress)

What's cool with Zooming User Interfaces is that you have always free space available anywhere (either by zooming or panning) to write new ideas.

That was the key idea in 2014 when creating BigPicture (ready-to-use infinite notepad in-the-cloud) and the open-source JavaScript library bigpicture.js powering it:

It works as expected on desktop browsers. Now, the next big challenge is: how to make it work on mobile devices?

It's funny to even have to ask this question, since touch devices are natively made to do panning (slide finger on screen) and zooming (pinch with 2 fingers). So it should be straightforward to adapt BigPicture to mobile devices.

However here are the difficulties:

  1. The transform/scale from CSS has limitations (probably max 10x or 100x factor when I started this project a few years ago), so we can't only use this to do a (nearly) infinite zooming user interface

  2. It requires to be able to zoom on a particular part of the viewport and not zoom the other parts of the HTML document (e.g. a top navigation header). Here are many potential solutions:

  3. Possible useful tools for this:

    • Zoomooz (however, I read in comments: Zoomooz does not support multi-touch pinching events. Its only a library for zooming into elements on a page, but has no support for pinching behavior, so far as I can see in the documentation.)

    • Hammer.js

    • ZUI53

    • TouchSwipe, a jQuery plugin for touch devices

Work in progress!

By the way, here is how to simulate touch events on Chrome for desktop computer: open the Developer console (F12), then there's a top-left button "Toggle device toolbar" (CTRL+SHIFT+M), here you go! For pinch-zoom events, use SHIFT + click + mouse up.

Your tests / pull requests / help to build a mobile version are welcome on this branch: https://github.com/josephernest/bigpicture.js/tree/mobile!

If you really like that open-source project, you can donate here: 1NkhiexP8NgKadN7yPrKg26Y4DN4hTsXbz.

Writing, a text-editor in the browser

Since I've started using StackOverflow, I've always loved their text editor (the one you use when writing a question/answer), because it supports Markdown syntax (a very elegant markup language to add bold, italic, titles, links, itemization, etc.), and even MathJax (which is more or less LaTeX syntax in the browser). I've always wanted to use such an editor for my own documents.

After some research, I found a few existing tools, but:

Let's go and actual build one! Here is the result, Writing:

Here's the source: https://github.com/josephernest/writing

For sure you'll like it!

If you really like that, you can donate here: 1NkhiexP8NgKadN7yPrKg26Y4DN4hTsXbz

Yopp — an easy way to send a file from phone to computer

Have you ever spent more than 1 second wondering:

"How do I get on my computer this photo I just made with my phone?"

or

"How do I get this PDF from my computer to my phone?"

Then you probably thought "Let's use Dropbox! ... oh no I'm not logged in on my phone, but what is my password again? Well, let's send the file to myself via email! Maybe I should just use a USB cable... but where is my USB cable again?"

Yopp is a solution for this problem, that you can easily install on your web server.

Thoughts about user experience & user interface design

This tool - Yopp - requires a total number of 7 actions to get the work done:

Open browser on phone [1 tap], Open Yopp page [1 tap if it's in the bookmarks], UPLOAD [1 tap], Choose file [1 action]

Open browser on computer [1 double click], Open Yopp page [1 click if in bookmarks], DOWNLOAD [1 click]

I'll be happy to switch to another tool if one requiring less actions exists.

I noticed that my likelihood/probability to use any tool (all other things being equal) is more or less proportional to P = 1 / a^2 (*) where a is the number of required actions/user inputs. If the number of required actions is doubled, the likelihood to use the tool is divided by 4.

Thus, even if it might sound obvious, one key element for a good user interface is to minimize the number of user actions to get a task done. If not, the user might unconsciously remember that the interface is unnecessarily complicated to use. He will then forget about the product, and look for another solution. (OK this is probably what will happen for you with Yopp if you don't have a web server already!)

As an example, I'm sure I'd use my city's bicycle sharing system Velo+ much more if I could take a bike by just swiping my card on the bike station's card reader (this is technically possible). Instead we have to: Tap on a screen (1), Choose "Subscribed user" (2), Swipe the card (3), Choose "Rent a bike" (4) (this one is particularly unuseful), Accept conditions already accepted many times before (5), etc. at the end it requires at least 12 actions! Any user who has done it at least once will process this data (required amount of inputs) and will probably make the choice of not using it for short distance trips.

It would be interesting to get more statistical data about the empirical result (*), this will be discussed in a future post.

See the BigPicture — a zooming user interface

This topic has been present in my thoughts for a long time, probably years:

“How to be able to think/write about lots of unrelated various topics, and still have a way to look at the big picture of what you’re doing?”

Here is my contribution about this:

  1. bigpictu.re, a ready-to-use infinite notepad (infinite zooming and panning)
  2. bigpicture.js, a JavaScript open-source library that you can use in various projects
  3. A standalone version of 1. (so you can take notes offline) is also available here: bigpicture-editor
  4. AReallyBigPage, an infinite collaborative notepad. It has been a real chaos once hundreds of people joined in. Probably internet’s deepest page ;)

 

Such an interface is called a Zooming User Interface (interesting reading: The humane interface by Jef Raskin, one of the creators of the Apple Macintosh), and strangely, ZUI has been very few used in modern interfaces.

As of 2017, nearly every software interface uses a 2D, or even a 1D navigation process: a web page only offers two scrolling directions: north and south. Even nowadays's apps famous for their "new kind of interface" still use a 1-axis navigation: "Swipe left or right".

Is there a future made of new interfaces?

TinyAnalytics

After having tested many open-source website analytics tool, and haven't found exactly what I was looking for, I started a minimalist project (coded in PHP) that only does this:

If you're looking for a tool lighter than Piwik, Open Web Analytics or Google Analytics, then TinyAnalytics might be what you're looking for.

Get the reverb impulse response of a church

I recently recorded an impulse response of the reverb of a 14th-century church (more or less the footprint of the sound ambiance of the building). Here is how I did it.

Quite a lot of reverb, that's exactly what we want to catch with an IR!

Then, of course, we can do some cleaning, fade out, etc.


But what is this useful for? You can use this Impulse Response in any music production software (the VST SIR1 is quite good and freeware) , and make any of your recordings (voice, instrument, etc.) sound like if they were recorded in this church. This is the magic of convolution reverb!


Useful trick when you record your own IR: play sweep0.wav in the building instead of sweep.wav. The initial "beep" is helpful to see exactly where things begin. If you don't do that, as the sweep begins with very low frequencies (starting from 20 Hz), you won't know exactly where is the beginning of your microphone-recording. Once your recording is done, you can trim the soundfile by making it begin exactly 10 seconds after the short beep.

Some related reading in this topic, and this blog post.

Older articles