An improvement of Shazam app to extend song identification beyond algorithms and involve people in the process. The app records all the songs that were tried to shazam by the user and keeps the recordings of those failed-to-shazam. These short sound samples are then played instead of ads on music streaming platforms for people to hear and efforts to identify the song are incentivized by offering rewards.
Identify failed-to-shazam songs.
Involve people in the process (fun+rewards).
Replace annoying sound ads on music streaming platforms with something better.
Collaboration between Shazam and music streaming platforms beneficial for both.
With this function in place, Shazam could enable identification of live played music, not only recordings.
Cases, when Shazam fails to identify a song, are quite frequent. If the song is not well known, it often takes many subsequent tries for it to finally get it. Many people don't have the patience for this. I think the main cause of this is that the app has to compare the sound input through the device's microphone to a huge number of songs in its database in a short period of time. Some lesser-known recordings might also not even be present in its database.
On the other hand, there's a high chance that someone knows the song Shazam has just failed to identify and could find the title/artist for you, it's just a matter of how many people you can show it to.
The app would record each song you're trying to shazam (through your phone's microphone) and if it happened to find the song, it would then delete the recording, if it didn't, it would keep it in phone's memory and later upload it to the cloud.
The app would use sound processing software to remove as much noise from the recording as possible and produce a sound sample focused on the melody of the song.
These short samples would then be played instead of sound ads on various music streaming platforms for users who use basic, unpaid version (e.g Spotify free) of the platform. Instead of a commercial, marketing a certain product, the user would first hear a voice encouraging to try identifying the sound sample, e.g.: "Can you tell what song is this?" and then the sample.
While the sample is playing, the icon saying "identify" or a similar one would appear on player's dashboard which you could click, type the name of the song and artist and submit to the system.
You could pause the sound sample "add" while it's playing and check the song you suspect it represents somewhere else on the web to make sure your guess is right.
To improve the chances of some users guessing the song correctly, the software could perform musical analysis of the recorded samples and adapt the sound "adds" to the taste of particular listeners. Spotify and other music streaming platforms already know what type of music each of its listeners is into, if the software finds a matching pattern between a failed-to-shazam sound sample and the musical preferences of particular listeners it plays that specific sample only to those listeners. This would be similar to ad personalization.
The motivation of users to identify sound samples would be incentivized by the music streaming platform(s), e.g. users who identified a certain number of songs would get a subscription of a Premium version of the app for free.
Shazam could make this extended identification feature a part of paid Shazam version. Part of the income from users paying the subscription would go to the music streaming platform(s) Shazam collaborates with to make this happen.
Confirming the accurate identification: This can be done in a couple of different ways - first, after the user submitted the name of the song and the artist to the system, the software would compare the recording sample to the recording of the suggested song existing anywhere online (in the same way Shazam compares the music heard through phone's microphone to the song recordings in its database, except this time the system would know which particular recording the sample should be compared to).
The second option is for a Shazam user who attempted to identify a certain song to get notified that someone has potentially identified their failed shazam. The user would check the proposed title of the song and then listen to it somewhere online and decide whether the proposition is accurate.
Both of these options could be used together as complementaries.
Privacy: As I mentioned, the software would try its best to filter out all the environmental sounds recorded during the shazaming process and leave only the music sample, but additionally, for privacy reasons (in order not to let masses of people hear accidentally recorded bits of conversations and other potentially sensitive data), before uploading the sample to the cloud, and playing it as an "add" on music streaming platforms, it could be mandatory for the user to give privacy consent.
The app would display a pop-up notification "please listen to this recording and make sure there aren't any private data in it that others can't hear". Only after the user press the "consent" button, the recording would be uploaded to the cloud.
Other remarks: Even the solitary function of Shazam keeping recorded sound samples in phone's memory for some time (without uploading them to the cloud and letting masses of people hear them) would be beneficial, because it would let the user play the recording for someone from the staff of the venue, the Dj after the performance, etc. and enable the chance for those people identifying the song by hearing a recorded part of it.