Music recognition app

Is there no museum, university or other special institution that has the full license and could provide an umbrella for the FOSS?
Are the hashes themselfs of non-free music non-free, too?
How about the software and starting with free and non-copyrighted software and people need to build up the database on their own by automatically downloading music from youtube and so on?

Good questions, have a lawyer at hand?

I sent a mail to licence-questions fsfe org.

I would donate in an eventual crowdfunding.

I got an autoresponder:

Organization: Free Software Foundation Europe e.V.
Thank you for your inquiry, which we will discuss internally and provide you with a response.
Please hold in mind that we cannot guarantee instant response times since our team consists of volunteers and license-related questions often require a complex discussion. However, we will do our best to respond within a reasonable time.

Thank you Jens. =)

I am thinking about how long do they need? When can I ask again? :wink:

For the full documentation I can paste the mail I sent:

Hi

At the F-Droid Forum “Hello” asked about a free Shazam app in the missing app thread. I started an extra thread [1] to not get offtopic.

Shazam FOSS would ne awesome.

Too complicated, hard to do without money since you’d need to
license all that music that you hash :frowning:

Is there no museum, university or other special institution that has
the full license and could provide an umbrella for the FOSS? Are the
hashes themselfs of non-free music non-free, too? How about the
software and starting with free and non-copyrighted software and
people need to build up the database on their own by automatically
downloading music from youtube and so on?

Can you answer some of the question?

Thanks

Jens Korte

[1] Music recognition app

It depends what you will do - only recognize and give name, artist, etc; or recognize then sell, or recognize then link to a streaming service, or…
Is it illegal to play music into a FOSS oscilloscope app https://f-droid.org/en/packages/org.billthefarmer.scope/ and see the “fingerprint?” Nah. Is it illegal to have a submit button for users to submit those “fingerprints” to your database? Probably not, but better get legal advice and form a shell company. Is it illegal to have a FOSS front-end for youtube? https://search.f-droid.org/?q=youtube&lang=en Nah. So is it illegal to link those together? Shouldn’t be… but sounds like $millions of work.

Wikipedia is full of errors but the history is interesting: https://en.wikipedia.org/wiki/Shazam_(application)

(IANAL)

I found: https://en.wikipedia.org/wiki/AcoustID

Maybe there is an Open Source support for music recognition in MusicBrainz or AcousticID but I didn’t find detailed information and there are different licenses for different data.

There is some edit support in (the desktop?) VLC which supports acoustic.org which I don’t know.

Other (maybe) useful links:

Database(s)

Content recognition

1 Like

Did someone already try https://github.com/frank240889/AutoMusicTagFixer ?

From the Readme:

For every audio file you have in your smartphone (all supported audio types by Android and jAudioTagger library), it will generate an audio fingerprint, then will connect to an ACR server and will download the correct data for every song, later will allow you to correct the song with this downloaded information; in other words, it will make a Shazam-like recognition but not from audio recorded, but over audio files you have and will correct their meta tags, including filename.

Edit: GPLv3

Encore (outdated)

Audio processing with python

alternativeto.net

I got a reply, but “we do not have neither capacity nor authorization to provide legal advice”, so don’t use this as your only advice how to write your free software. In an extra mail I got the permission to publish this mail.

Dear Jens,

thank you for contacting us and sorry for the delay in the response. Momentarily we are overloaded with requests. Please note that we do not have neither capacity nor authorization to provide legal advice. Although we can give you some pointers, please understand this is not a legal advice.

From our perspective, the use of the hashes that are created depend on the technology that is used. For instance sound recognition could be of various types - fingerprinting/tagging, real-time identification, OTA (Over-The-Air) recognition, etc. These hashes do not operate like APIs. Hashes are unique identifiers stored in the database after audio fingerprinting generated by using an algorithm (proprietary in the case of Shazam).

Shazam has a database to match against the fingerprints of the audio clips by way of music distribution, whereby record companies provide songs to Shazam to analyze and build database. Shazam has also partnered with famous labels, music, television and advertising companies, movie makers and even artists directly. Therefore, since Shazam’s fingerprinting technology is proprietary, the resulting hashes are also ‘closed’.

As in thread mentioned by you, there are Free Software options available for this technology. In principle, if the audio fingerprinting technology is open, then the hashes created by such method should also be open.

On another note, another factor to consider is the copyright issues emanating from the music distribution and database creation. All songs are owned by the copyright holders (e.g., SONY, EMI, etc.), however they license the same to Shazam for creation of database and derive royalty from the same. As regards, creating a library from royalty-free music is concerned, each royalty free music company works differently so it’s important to be aware of terms and conditions before using their services.

Hope that helps,

Kind regards,
Lucas

It would be awesome if there were a foss alternative to shazzam.

I wonder how well https://acoustid.org/ fingerprinting would work in a song recognition app. (It’s the back-end for music brainz, vlc, etc.) I don’t think they ever ran into legal issues. In their usecase however they never have to deal with poor audio snippet quality so their solution will likely perform bad in a music recognition app.

It’s not directly usable for this, because it was not designed for it. AcoustID, the actual server side lookup service, was created to identify whole audio file, and intentionally holds only fingerprint for the first 60 seconds of a song. I can’t find the source right now, but Lukas Lalinsky, who created and operates AcoustID, once mentioned that this restriction even made it possible for him to operate this on an affordable basis, because it limits the required data storage and processing power.

The other part is Chromaprint, the fingerprinting algorithm. It also was designed to identify clean audio, and it does not work well with very short snippets.

The following mailing list discussions have some details about this:

I think the issues are probably not so much legal ones. A database of fingerprints likely has no legal problems. There might be some patent issues, but that’s something we have in other parts of FOSS software as well.

The technical side should not be underestimated. But as indicated by the answer from the FSF there are existing open source audio fingerprinting solutions, and there is extensive audio metadata on e.g. MusicBrainz. I’m sure someone could come up tying this together for a FOSS audio lookup server.

But the biggest challenge then probably is to operate such a service and gathering the data (fingerprints). Full audio fingerprinting surely uses significantly more storage compared to what AcoustID has right now.

1 Like