Now We Know What This Song Is. SoundHound vs Shazam
SoundHound is the obvious winner for many practical reasons. Say you want to sing along to a tune you hear in the grocery store (with noisy shoppers and their kids bustling about); you pull out SoundHound, it automatically begins listening for the tune. 5 to 15 seconds later, you have the lyrics right in front of you and you begin to sing along. …And she smiles an says, oh you know that song too? Or for instance a TV commercial in the living room with the air conditioner on; the song is Sooooo Awwwesome! you pull out SH, grab it, bookmark it, and now forever have the YouTube link to listen to it at your pleasure. Thats how I got the rad song from those Escalade commercials last year “Clubbed to Death”. Etc. Etc. Etc. SoundHound wins ‘hands up, hands up’.
An awesome comment to some comparative review
Most reviews of these apps on the Web start with the situation modeling, and I won't stand out. So, you are at sort of a party. I hope it looks something like this:
And at this moment, of course, you’ve decided to find out the title of the unbelievable song that charms your ear.
Then everything is pretty much simple: you remember few refrain’s words (it is not a challenge since nowadays refrains are usually few-word-pieces). Go to Google, type them + the word “lyrics”. And voila! Most likely during the check of the first result you’ll see something like this:
Works in 95% of all cases.
** Voice of the editor Alex, who has sneaked up on me**
-Holy crap! What are you talking about, what Google? You was supposed to write about the apps recognizing music song aurally, wasn’t you?
- Damn.…
**
Umm, let's take few steps back. Specifically for such kind of situation there are the mobile apps Shazam and SoundHound.
“Recognition of a musical piece based on its fragment recorded via microphone” – sounds too complicated. That’s why they found a better fitting word – a tag. The process of recognition is called tagging.
How do they work?
In a nutshell, if to be serious about it, there is a reasonable, detailed and understandable review over here http://laplacian.wordpress.com/2009/01/10/how-shazam-works/. I only mention that they don’t send anywhere the sound stream you directly fed to the app – at first they process it on a phone. If to put it really simple it works in such way:
1. At first, a sort of an imprint of your music piece is indented (just like taking human’s fingerprints). Imagine it as a table where one column shows the time from the start of the record, and the second column shows frequency (in Hz) currently being at the peak of intensity.
2. Then the app sends the imprint to the server that compares it with the database of other imprints, stored as files on the numerous music tracks.
3. The title of the track is sent back from the server to the phone.
Such imprint is called a “spectrogram”. Google since it is interesting.
How do they look like?
Shazam
SoundHound
It’s easy to see with a half an eye, that the interface of SoundHound was created in a more professional way. For example, the process of recognition looks like this:
Small white clouds and yellow rays in SoundHound are dancing the hell out of it to the music– fucking awesome!
Shazam from time to time displays the ad of the advanced app’s version, and that is as effective as a phone call interrupting sex.
It is more “delicious” and delightful to use SoundHound, gets a plus for this, though the factor isn’t decisive.
But there is a couple of really solid arguments. The free version of SoundHound has a set of more powerful functions. For instance, after tagging SoundHound shows not just the title and the artist of the track, but also streams the song’s lyrics against the background of the album’s cover displayed. Neat!
Besides, there is a built-in audio player. And it is also great, and later I'll write why.
Shazam has all of this too, but, firstly, only in the paid version, and, secondly, a bit incompletely – as two products, LyricPlay and Shazam Player. The free basic version is quite modest and sad-looking:
The paid version of SoundHound for $7 doesn't bother with ads but does gladden with few “cookies” like music recommendations.
The main thing is that now the free versions of both apps have no the limit for the songs to tag. And there’s been the limit earlier. This is a big minus for the users of the previous paid versions, since they had to pay for the limitless now apps that used to have limits. But for others it's a big plus.
Testing
In general, theoretically, there is a plenty of situations when these apps can be very appreciated.
After scratching my head for a while, I divided them into two categories:
1. Simple. With the low level of the outside sounds.
For example, if you are driving and listening to the radio or your home sound system is playing an unknown Track 12. In such cases the app's ability to recognize a song rather depends on the size of its database than on its algorithms. Granted, if you are not crunching chips too loudly.
The database used by SoundHound is http://midomi.com/. You can test its service directly from the site by singing something into your laptop's microphone.
On the Shazam’s site - http://www.shazam.com/ - nothing interesting was found.
2. Complicated. With the high-noise outside sounds.
Among people at a supermarket or in public transport; or the road noise if you are transported in a car trunk.
In this case the impact of these noises on the tagging will play a significant role.
I tested far and wide both apps in different variations of these two categories. Below you can see what results the tests brought.
Simple
App Total tries Successful
Shazam 50 40
SoundHound 50 32
Complicated
App Total tries Successful
Shazam 20 11
SoundHound 20 13
Interpretation
It is worth saying a few words about the method of testing. I tried not to run an experiment with a spherical horse in vacuum, so to speak, that’s why I just used the apps in everyday life as much as possible. I ran into different stuff – popular and rare music, a bunch of various genres, remixes and covers. So, what can I say?
1) On conditions of the outside noises’ absence, Shazam easily beats SoundHound. The database is absolutely huge. It happens when a really rare track is playing and you start thinking with a malicious smirk: “Nah, this one you're not gonna find, for sure”. But Shazam finds it! I get a feeling that if I hum a song by accident in 5 minutes it will be in its database.
2) SoundHound works mostly with the English language content. The Shazam database contains music from a great amount of countries. Such music that even in these countries is considered as rare and unpopular.
3) On the really lousy conditions with the high-noise level of the outside sounds, both programs work in a lame way. Well, it is so very understandable. Nevertheless, SoundHound seemed to work better, as for me.
4) And the last observation, SoundHound works about 1.5-2 times faster. It sends the imprint to the server not for once (at the end of the record like Shazam does) but more frequently. Sometimes it recognizes a song just in few seconds! Faster than I do!
Hardcore testing
“Just sing or hum a song”, - proudly announces SoundHound, hinting at the significant improvements of recognition algorithms comparing to Shazam. Sounds cool. Well, let's check. Not every day someone says I sing like Sinatra.
Not that this time anyone did it either. So let's feed SoundHound with the little-known covers of the known to everyone song by Adele - Rolling In the Deep. The covers are not in its database.
An awesome instrumental version by Piano Guys
http://www.youtube.com/watch?v=lUjWJSnGVB0
A slightly adolescent cover from three girls in many-colored U-neeq dresses. I fell for the one in a blue dress.
http://www.youtube.com/watch?v=Uu-fZvwfc9U
Another instrumental cover, this one is by Jun Sung Ahn. Great, but I found Piano Guys to be more interesting.
http://www.youtube.com/watch?v=K-eZfCs8yOc
A family version. The little girl’s dancing in a Beyonce-style.
http://www.youtube.com/watch?v=JtawDJtcRg8
Too wimpy, but great.
http://www.youtube.com/watch?v=Nd7ofygACcQ
Not hard to guess that it couldn't recognize Adele in the instrumental covers. But it did tag all the others (though not at the first try). So hum and sing. It really works.
Shazam became a nice surprise too since it has more than a half of these records in its database.
BTW, I’d like to know- what cover did you like most?
Conclusion
Nevertheless, all of this is just a pseudo-practice. In all conscience, I am sure that in real life most people install these apps just to try them and say, “WTF, it really works, fuck me!”
Then they tap the screen few times more and close the app with the feeling of the mission being accomplished and complete acknowledgment of the new heights of the technological progress.
In real life, this monkey trick can become useful maximum few times a year. And even then, according to the Murphy's Law, it won't work or a song will end before you have tap Listen.
But! Nevertheless, SoundHound and Shazam are genius applications. The genius programs with the most idiotic targeting.
They are not taggers. They are the world’s best foreign languages trainers. Here is a situation. Let's say, fro example, you learn Spanish. Excellent! Just download Shakira's songs to your iPhone, listen to them with the built-in player and read the lyrics while listening! During the study of any language audio perception is a crucial moment. Mix work and pleasure.
The bottom line is in the form of the answers to FAQ.
What app recognizes music best?
Shazam “knows” more music, SoundHound “hears” it better. Both apps are good.
What about the additional functions?
In general, they are similar, but SoundHound can offer much more in the free version.
What app should I pick? A free or paid version?
Install both apps in free versions. I don’t find sense in buying them.
P.S.
We figure out what ear for music both apps have, but what kind of hearing do you have? While I was researching for the article I found simple and interesting tests that can tell you the stuff you already know.
The test of tone difference perception – measures the ability to distinguish sounds of different frequency.
Test of tunefulness – measures the ability to distinguish melodies. You have to define whether two melodies are the same or different.
Rhythm test – measures the ability to distinguish rhythm patterns.
