Hi Viniciusccarvalho - You can get analysis data for any of the many millions of songs that we've already analyzed. If you have a track ID (such as a Rosetta ID from Spotify, Rdio or 7Digital) you can get the analysis data for that exact audio like so:
track/profile with a 7digital-US track ID
The returned audio summary has an analysis_url that points to the full detailed analysis.
If you have the song artist and title you can similarly get the analysis. Of course, if there are many versions of a song we are likely to return an audio summary and analysis for a different version than the one you have. Here's an example of getting the audio summary given an artist and title:
Song search for Karma Police
Again, the returned audio_summary has an analysis_url that points to the detailed analysis.
If you want to make sure that the analysis is an exact match for your audio, you have two options. First, query for the analysis using an MD5 of the audio file,like so:
query by md5
However, this is a pretty unreliable approach since we internally only map one MD5 to an audio track, where in reality, due to encoding errors, bit errors and metadata differences, a single track could have many possible MD5s.
The better way is to use audio fingerprinting. You can use ENMFP or Echoprint to generate an audio fingerprint for each of your tracks. You can then query for the song details with the song/identify query.
Hope this helps.