Hermann Redies and the folks at Xeno-Canto have just launched an ambitious project called Pai-Luiz, which attempts to automatically identify recordings of unknown bird sounds by looking through the entire Xeno-Canto database for matching syllables.  It’s just a prototype system at the moment — not particularly user-friendly yet, nor particularly accurate, but it still represents a huge leap forward in online automatic sound identification.

To give Pai-Luiz a try, you have to log in as a Xeno-Canto user, upload two different WAV files of the sound, and specify a precise bandwidth — as Hermann explains in the online documentation, you need to be pretty familiar with sound editing and spectrograms in order to do all this.  Once you upload a sound, Pai-Luiz takes up to an hour to process your request and email you a long list of recordings that might match yours.  The list is only a group of best-guess suggestions and there’s no guarantee that it contains a match — the actual identification still falls to you, the human user.  But when it works correctly, Pai-Luiz cuts down a lot on your workload — instead of having to listen to tens of thousands of recordings, you only have to sort through a few dozen possible matches.

Hermann is looking for users to try out the system and give feedback so that he can improve it.  If you’ve got a little extra time, you might want to feed it some known and unknown sounds to see what it kicks back.

(By the way: Hermann is a co-founder of Association “Mãe-da-Lua”, which purchased a Nature Reserve for the birds of the threatened caatinga habitat in northeastern Brazil, but can no longer afford to keep it open.  The reserve is looking for buyers or donors; for more information, see Hermann’s website.)