How Podcasts Have Changed in Ten Years: By the Numbers '-- Medium
Sat, 05 Sep 2015 20:25
WE'VE ALL HEARD that a podcasting boom is underway. Since January 2013, the number of podcasts listed on iTunes US has doubled (see Activity below). Part of the allure for content producers is that podcasting is a wild frontier in the new media landscape, waiting to be shaped and settled. Some have questioned if podcasting is the new blogging, referring to a similar boom during the early and mid-2000's, but that remains to be seen.
I've noticed since starting a podcast of my own that research on the field is scant. Most of the research I've read has focused on listener behavior, which is fine for marketers, but other questions about the medium have gone unanswered. I decided to address a few.
What iTunes categories have the most podcasts?How many podcasts are launched per month?How many podcasts are active?How long is a typical podcast episode? How often is a typical podcast updated?How many podcasts have explicit content?How many podcasts are not in English?How many ratings or reviews does a typical podcast receive?I was also curious if these varied by category, e.g. ''Are episodes in some categories longer than others?'' or ''Which categories have more explicit content?'' Here was my approach.
MethodologyPodcasting is a young and decentralized medium. Thus far, Apple has been its most influential actor, having celebrated ten years of podcasts on its iTunes service in June 2015. During that span, Apple established many of the frontier's early boundaries, like branding the medium with its generic trademark: iPod and ''casting.
''ITunes 12 logo'' by Apple Inc. Licensed under Public Domain via CommonsDespite its dominance, iTunes does not house all of the world's podcasts. Some podcasters avoid iTunes or do not configure their shows for it. Apple also separates its iTunes Store by country, so many international podcasts are absent from the US Store. Still, iTunes US was the best place to begin a sketch of the medium because its podcast directory was the largest at the time.
Because Apple has not published a dataset on iTunes, I had to construct my own. I started by copying the names of all podcasts on the iTunes US directory as of June 2015. I divided them by iTunes' categories and counted the names in each category. From there, I randomly selected a sample of 100 each from the twenty-five categories with the most names, which I combined for a total sample of 2,500.
I gathered details on all 2,500 podcasts in the sample by searching for them in iTunes 12. I scanned the profiles of each for values of interest between June 2005 and June 2015: activity status; start date (and end date, if applicable); months active; items per month; Explicit status; English or non-English descriptors; median episode times; and number of ratings. Many of these were calculated using spreadsheet formulas.
Here's what I came up with. The data are available for categories, for cumulative podcasts, activity, and launches, and for language, explicitness, and minutes per episode. Charts were assembled in Excel and Photoshop.
ResultsCATEGORIES: What iTunes categories have the most podcasts?The top three categories by quantity: Christianity, Music, and Comedy.
In June 2015, the directory for iTunes US listed over 271,000 podcasts. There was a catch: roughly a quarter of these were listed under multiple categories. For example, the most abused category was Podcasting, where about 11,500 of the 15,000 listings did not list Podcasting as the primary category''--'they were duplicated from other categories.
After adjusting for duplicates, I calculated that the true number of podcasts on iTunes US as of June 2015 was around 206,000. I also measured the twenty categories on iTunes US with the most podcasts.
The categories with the fewest podcasts were: Aviation (150); Hinduism (260); and Regional, a subcategory of Governments & Organizations (320).
ACTIVITY: How many podcasts are active?About 40 percent. Over 60,000 per month were active as of 2015.
Most podcasts are short-lived. Many go dormant after their hosts complete goals or lose interest. Others are part of larger collections, like audiobooks based on classical texts, and are left alone once they are listed.
Between June 2005 and June 2015, a typical podcast ran for six months and twelve episodes, at two episodes per month, before going inactive.
For this research, active podcasts were measured as those with at least one new item between January and June of 2015. I estimated the number of podcasts fitting this criterion on iTunes US at about 40 percent of the total, or roughly 84,000 of the 206,000 podcasts. The chart below shows that, as of June 2015, about 60,000 podcasts were active per month.
The categories with the most podcasts active in 2015 by total:
Christianity: 23,200 active podcastsMusic: 13,100 active podcastsComedy: 7,300 active podcastsTV & Film: 6,400 active podcastsNews & Politics: 6,300 active podcastsThe categories with the most podcasts active in 2015 by percentage:
News & Politics: 61% of podcasts in the category were activeChristianity: 59% activeProfessional (Sports & Recreation): 53% activePhilosophy: 52% activeComedy: 51% activeThe least active categories by percentage were Podcasting (23% active), K-12 (25% active), Literature, and Visual Arts (tied at 29% active).
LAUNCHES: How many podcasts are launched per month?In 2015, podcasters added about 5,000 new podcasts to iTunes US per month.
The chart below shows a surge in new podcasts beginning in late 2013. Due to erratic results, I found a trendline to be informative. (The upward trendline does not guarantee that the launch rate will continue to increase.)
One indicator of a podcasting boom was that nearly a third (30%) of all podcasts listed on iTunes US were launched between June 2014 and June 2015. The categories with the most launches during that 12-month period:
News & Politics: 3,700Professional (Sports & Recreation): 3,100Christianity: 3,000Self-Help: 2,800(tie) Management & Marketing, Philosophy: 2,600TIME & RELEASE RATE: How long is a typical podcast episode? How often is a typical podcast updated?As of 2015, a typical podcast published two 40-minutes episodes per month.
The median number of minutes in a typical podcast episode has increased, from 25 minutes in June 2007 to 40 minutes in June 2015. See below.
The categories with the longest median times per episode:
Video Games: 52 minutesPhilosophy: 51 minutes(tie) Music, Spirituality: 50 minutesProfessional (Sports and Recreation): 48 minutesThe categories with the shortest median times per episode were K-12 (10 minutes), Podcasting, and Business News (tied at 15 minutes).
A preliminary analysis suggested that the median release rate increased from less than two episodes per month in 2007 to the semi-monthly rate, and it may still be increasing.
EXPLICITNESS: How many podcasts have explicit content?As of 2015, about four in 25 podcasts were tagged Explicit at least half of the time.
When listing their podcasts on iTunes, podcasters can tag their episodes as either Clean or Explicit. Assuming the podcasts in the sample were tagged correctly, the percentage with over half of their episodes tagged Explicit almost doubled, from 8.4 percent in June 2007 to 16.6 percent in June 2015.
The most explicit categories by percentage:
Comedy: 56% explicitVideo Games: 39% explicitHobbies: 28% explicitPhilosophy: 26% explicitPerforming Arts: 25% explicitFive out of the 25 categories had no explicit podcasts in their samples: Business News, Christianity, K-12, Kids & Family, and Self-Help. That's not to say explicit podcasts don't exist in these categories, but they appear to represent less than one percent of the podcasts in each.
LANGUAGE: How many podcasts are not in English?As of 2015, roughly one in five podcasts on iTunes US were not in English.
The United States is becoming more racially and ethnically diverse. So are the podcasts on iTunes US, at least linguistically. Non-English podcasts gained about ten percent on English podcasts between 2007 and 2015; by June 2015, non-English podcasts comprised 22 percent of the sample.
The categories with the most non-English podcasts by percentage:
(tie) History, Philosophy: 36% non-EnglishNews & Politics: 33% non-English(tie) Places & Travel, Podcasting: 32% non-EnglishThe categories with the fewest non-English podcasts by percentage were Performing Arts (9% non-English), Comedy, and Self-Help (tied at 11% non-English).
USER RATINGS: How many ratings does a typical podcast receive?Most receive none. Among those that do, seven was par for the course.
On iTunes US, users can sign in with their Apple accounts and leave ratings for podcasts. They also have the option to leave written reviews.
The median for ratings was a goose egg, as only 24 percent of podcasts in the sample had at least one rating. Among those, the median was seven.
Alda: A Manifesto and Gentle Introduction :: Dave Yarwood
Sun, 06 Sep 2015 15:19
date:September 5, 2015
Alda's ambition is to be a powerful and flexible music programming language that can be used to create music in a variety of genres by typing some code into a text editor and running a program that compiles the code and turns it into sound. I've put a lot of thought into making the syntax as intuitive and beginner-friendly as possible. In fact, one of the goals of Alda is to be simple for someone with little-to-no programming experience to pick up and start using. Alda's tagline, a music programming language for musicians, conveys its goal of being useful to non-programmers.
But while its syntax aims to be as simple as possible, Alda will also be extensive in scope, offering composers a canvas with creative possibilities as close to unlimited as it can muster. I'm about to ramble a little about the inspiring creative potential that audio programming languages can bring to the table; it is my hope that Alda will embody much of this potential.
At the time of writing, Alda can be used to create MIDI scores, using any instrument available in the General MIDI sound set. In the near future, Alda's scope will be expanded to include sounds synthesized from basic waveforms, samples loaded from sound files, and perhaps other forms of synthesis. I'm envisioning a world where programmers and non-programmers alike can create all sorts of music, from classical to chiptune to experimental soundscapes, using only a text editor and the Alda executable.
In this blog post, I will walk you through the steps of setting up Alda and writing some basic scores.
But first, a little history.
I can trace the origins of Alda back to 2004. At the time, I was studying Music Composition at UNC and exploring ways to make electronic music. I mucked around in FruityLoops a bit without much success. Eventually, I stumbled upon an entire world of music programming. My gateway drug was MML, which was (and perhaps still is) the most legit way for chiptune musicians to make NES music. After reading through Nullsleep's excellent MML tutorial and learning how to make some rudimentary NES music, I started to become more involved in making music with human beings and took a hiatus from MML. But the seed had been planted for a lifelong obsession with the idea of programming music.
MML ended up becoming a major influence on Alda. I really enjoyed the workflow of creating NES music by writing code in a text editor. But what I wanted was a more general-purpose music programming language. I wanted to take MML's approach to generating NES music and extend it to other realms of digital music creation: additive/subtractive synthesis, electroacoustic music, and even classical music.
I was a classically trained musician long before I was a competent programmer. My music education led to a particular interest in composing music in a variety of styles. Growing up around computers, I discovered an ever-expanding class of GUI applications designed to help musicians compose music. I wrote guitar tablature with Guitar Pro, and for traditional music notation I have tried, at various times in my life, Cakewalk, Noteworthy Composer, Finale, Sibelius, and MuseScore, among others.
When I was studying music composition, I got a lot of mileage out of Sibelius, in particular. I used Sibelius extensively to transcribe pieces of music as I composed them. It allowed me to have a digital record of my compositions, and it was essential (practically a requirement) as a way to print out individual instrument parts to distribute to the musicians who performed my pieces.
Music notation applications like Sibelius are clearly a very important tool for people who are serious about composing music. However, as a programmer and as a composer, I feel that there are a couple of fundamental problems with GUI music notation software:
It's distracting. When pre-digital age composers used to write music, they would sit at a piano and write it out by hand on staff paper. All of the notation techniques, the layout of their scores, everything, came directly from their minds and through their pens. When you notate music using a GUI application, you have menus upon menus in front of you from which to select whatever elements of music notation you wish to use in your score. This is distracting in two ways:
It's not always easy to find what you're looking for. By the time you find it, you may have lost your train of thought.
Having all of these music elements in front of you is visually distracting. When I used Sibelius, I would sometimes end up distracting myself for long stretches of time as I perused all of the different music notation elements it was possible to place, or browsed through all the project templates available.
In contrast to working with these complex GUI applications, I have found that programming pieces of music in a text editor is a pleasantly distraction-free experience.
It's limiting. I think for a composer to have an ideal environment in which to compose, he needs to get back to the basics. He needs a blank canvas and a way to notate music. And because this is the 21st century, his scores need a way to be interpreted by a computer and turned into audio. It would also be nice if his scores could be easily converted to and from the standard notation format that human musicians are trained to read.
The GUI programs available today do an excellent job of handling these 21st century requirements, but they do so by taking a shortcut '' they skip the ''blank canvas'' part. Of course, when you create a new score in Sibelius (for example), you do have what looks like some empty lines of staff paper, but in fact, this blank staff paper carries a very different connotation than does a physical page of manuscript paper. You can't just grab a pencil and start writing whatever your heart desires. There are a number of hidden restraints that the GUI application forces upon you.
This is an inherent shortcoming of any GUI music notation editor; in order to be able to represent your musical score visually in a sane and comprehensible way, it has to impose some restrictions. Audio programming languages must also impose restrictions (in the same sense that any piece of software does), but because they are not tied to visually representing your score and maintaining a user-friendly GUI interface, they are able to get away with imposing substantially less restrictions on the composer. As a composer myself, I find this fascinating and inspiring.
Note: As Alda is currently still under development, you will need to follow the process below in order to run it. Hopefully the process is pretty intuitive, but please feel free to e-mail me if you run into any issues. In the future, the process will be much simpler, e.g. downloading and running a standalone executable program.
To get started with Alda, you will need two things:
The Clojure build tool Boot. If you are running Mac OS X and you have Homebrew installed, you can install Boot by running brew install boot-clj. Otherwise, see here for more details about installing Boot.
The alda executable file, which you can install by copying and pasting the following command into your terminal:
curl https://raw.githubusercontent.com/alda-lang/alda/master/bin/alda -o /usr/local/bin/alda && chmod +x /usr/local/bin/aldaYou should now be able to use a handful of built-in commands that start with alda. You can parse and/or play Alda code from a file or a string of Alda code provided as a command-line argument. Or, you can build a score incrementally by using the Alda REPL (Read-Evaluate-Play Loop).
We will use the Alda REPL at first, to experiment a little with Alda syntax. To start the REPL, type:
Note: Alda uses a soundfont called FluidR3 to make MIDI sound a lot nicer. This is a one-time 125 MB download that will kick off the first time you run the above command. This may take a few minutes or longer, depending on your network connection. To pass the time while you wait, you may want to watch somemusicvideos on YouTube or something.
If you're feeling impatient and you'd like to skip this step and use the Java Virtual Machine's built-in MIDI synthesizer (which sounds terrible) instead, you can type alda repl --stock.
Once FluidR3 has downloaded and the REPL is ready, you should see something like this:
You can type snippets of Alda code into the REPL, press Enter, and hear the results instantly.
As I mentioned, MML ended up being a primary influence on Alda, along with LilyPond. The great thing about both MML and LilyPond is the simplicity of their syntax. I would describe the syntax of both languages as being similar to Markdown; essentially, what you see is what you get.
NotesLet's start with a simple example. Let's translate this measure of sheet music into Alda code:
Here we have four quarter notes: C, D, E and F. The Alda version of this is:
Try typing this into the REPL and pressing Enter'... nothing happens. Why? Well, we haven't told Alda what instrument we want to play these notes. Let's go with a piano:
Now you should hear a piano playing those four notes. You will also notice that the prompt has changed from > to p>. p is short for piano, and it signifies that the piano is the only currently active instrument. Until you change instruments, any notes that you enter into the REPL will continue to be played by the piano.
OctavesLet's add some more notes.
You should hear the piano continuing upwards in the C major scale. An interesting thing to note here is the >. This is Alda syntax for ''go up to the next octave.'' An octave, in scientific pitch notation, starts on a C and goes up to a B. Once you go above that B, the notes start over from C and you are in a new octave. In Alda, each instrument starts in octave 4, and remains in that octave until you tell it to change octaves. You can do that in one of two ways: you can use to go down or up by one octave; or, you can jump to a specific octave using o followed by a number. For example:
o0 c > c > c > c > c > c > c > c > c > cAccidentalsSharps and flats can be added to a note by appending + or -.
You can even have double flats/sharps:
As a matter of fact, a note in Alda can have any combination of flats/sharps. It usually isn't useful to use more than 2 sharps or flats (tops), but there's nothing stopping you from doing things like this:
The above is a really obtuse and unnecessary way to represent an E (a.k.a. a C-sharp-sharp-sharp-sharp-flat-sharp-flat-sharp-flat-sharp) in Alda.
Note lengthsBy default, notes in Alda are quarter notes. You can set the length of a note by adding a number after it. The number represents the note type, e.g. 4 for a quarter note, 8 for an eighth, 16 for a sixteenth, etc. When you specify a note length, this becomes the ''new default'' for all subsequent notes.
o4 c4 c8 c c16 c c c c32 c c c c c c c | c1You may have noticed the pipe | character before the last note in the example above. This represents a bar line separating two measures of music. Bar lines are optional in Alda; they are ignored by the compiler, and serve no purpose apart from making your score more readable.
RestsRests in Alda work just like notes; they're kind of like notes that you can't hear. A rest is represented as the letter r.
Dotted notesYou can use dotted notes, too. Simply add one or more .s onto the end of a note length.
trombone: o2 c4.. d16 e-8 r c rTiesYou can add note durations together using a tie, which in Alda is represented as a tilde ~.
ChordsWhen you play multiple notes at the same time on a single instrument, that's a chord! In Alda, a chord is multiple notes separated by slashes /.
Notice that, just like with a sequence of consecutive notes, specifying a note length on one note of a chord will make that the default note length for all subsequent notes.
A convenient feature of Alda is that the notes in a chord do not need to be the same length. This can be convenient when writing pieces of music that feature melodies weaving in and out of chords:
o4 c1/e/g/>c4 < />eAlso note that it is possible to change octaves mid-chord using . This makes it convenient to describe chords from the bottom up or top down.
VoicesAnother way to represent notes played at the same time in Alda is with voices. The same example we just wrote with chords could also be written like this using a combination of chords and voices:
V1: o5 c4 < >To exit the Alda REPL, type bye and press Enter.
So far, we have been feeding Alda some code, line by line, and hearing the result each time. This is a good way to test the waters and see how small pieces of code sound before you commit to them. When you're ready to set some music down in stone, it's time to write a score.
In Alda, a score is just a text file. You can use any text editor you'd like to create this text file. By convention, the file's name should end in .alda. Create a blank text file in whatever directory you're currently in in your terminal, and name it test.alda.
Type the following into test.alda:
bassoon: o2 d8 e (quant 30) f+ g (quant 99) a2Then, run alda play --file test.alda. You should hear a nimble bassoon melody.
AttributesYou may have noticed that I snuck in a new syntax here. I was going to get to that, I promise! (quant XX) (where XX is a number from 0-99) essentially changes the length of a note, without changing its duration. The number argument represents the percentage of the note's full length that is heard. Notice, when you play back the bassoon melody above, how the F# and G notes (quantized at 30%) are short and staccato, whereas the final A note (quantized at 99%) is long and legato.
quant (short for quantization) is one example of an attribute that you can set within an Alda score. volume is another example; it lets you set the volume of the notes to come. Like most attributes, volume (which can be abbreviated as vol) is also expressed as a number between 0 and 100. You can set multiple attributes at once by separating them with commas.
Try editing test.alda to look like this:
bassoon: o2 d8 e (quant 30, vol 65) f+ g (quant 99) a2Run alda play --file test.alda again to hear the difference in volume between the first two and last three notes.
As an aside: you may have noticed there is a little bit of a wait every time you run an alda command. This is an unfortunate side effect of Alda being a Clojure project; Clojure has a notoriously slow start-up time. You might prefer to use the Alda REPL to experiment '' once the REPL is started up, you will hear the results of each line of code instantly.
It's also worth noting that in the near future, Alda will be available as an ahead-of-time-compiled, standalone executable, which should speed things up a bit.
Multiple instrumentsFinally, we come to the meat of writing a score: writing for multiple instruments.
An Alda score can contain any number of instrument parts, which are all played simultaneously when the score is performed. Try this out in your test.alda file:
trumpet: o4 c8 d e f g a b > c4.trombone: o3 e8 f g a b > c d e4.The key thing to notice here is that we have written out individual parts for two instruments, a trumpet and a trombone '' one after the other '' and when you play the score, you will hear both instruments playing at the same time, in harmony.
You can also write out the parts a little at a time, like this:
trumpet: o4 c8 d e f g trombone: o3 e8 f g a b trumpet: a b > c4.trombone: > c d e4.Notice that this example sounds exactly the same as the last example. This demonstrates another important thing about writing scores in Alda: when you switch to another instrument part, the instrument part you were working on still exists, in sort of a ''paused'' state, ready to pick it back up where you left off once you switch back to that instrument.
Global attributesRecall that you can change things like an instrument's volume by setting attributes.tempo is another thing you can change by setting an attribute. Let's try it:
trumpet: (tempo 200) o4 c8 d e f g a b > c4.trombone: o3 e8 f g a b > c d e4.Wait a minute'... did you hear that? The trumpet took off at 200 bpm like we told it to, but the trombone remained steady at the default tempo, 120 bpm! This is actually not a bug, but a feature. In Alda, tempo (along with every other attribute) is set on a per-instrument basis, making it entirely possible for two instruments to be playing at two totally different tempos.
Global attributes are written just like regular attributes, but with an exclamation point on the end. Try this on for size:
trumpet: (tempo! 200) o4 c8 d e f g a b > c4.trombone: o3 e8 f g a b > c d e4.tempo! sets the tempo for all instruments, at the specific time in the score where you place it. Try playing around with this bit of Alda code, moving the (tempo! 200) to different places in the score. Try out some different tempos other than 200 bpm.
MarkersWe've already gone over a lot, but I'd like to show you how to do just one more thing in Alda '' it's an important one because it helps you keep your instruments synchronized in perfect time.
The concept behind markers is assigning a name to a moment in time. A name can contain letters, numbers, apostrophes, dashes, pluses, and parentheses, and the first two characters must be letters. The following are all examples of valid marker names:
chorusvoiceInlast-noteverse(2)bass+drumsUsing markers is a two-step process. Place a marker by sticking a % before the name, and then jump to it by sticking a @ before the name. To demonstrate, let's go back to our trumpet and trombone example. Let's have a tuba come in right on the last note. We can do that by placing a marker in either the trumpet or trombone part, right before the last note, and then jump to that marker in the tuba part that we'll create:
trumpet: o4 c8 d e f g a b > %last-note c4.~2trombone: o3 e8 f g a b > c d e4.~2tuba: @last-note o2 c4.~2So, that's Alda in a nutshell. Please don't hesitate to e-mail me if you have any questions about how to do something in Alda. Or, better yet, if you're a Clojure programmer and you like open-source software, consider contributing! Pull requests are warmly accepted.
blog comments powered by