Audio Pitch Tracer was developed to transcribe musical pitches in microtonal music and bird song. It can capture pitches of very short notes (even sub-millisecond), and it captures the pitch to within a thousandth of a semi-tone for notes of a fifth of a second or longer. This means that even the minutest nuances of pitch can be transcribed, and Audio Pitch Tracer is therefore of interest to bioacoustic naturalists as well as to musicians and composers.
Audio Pitch Tracer creates a MIDI file from bird song recordings or any other clean recordings of a solo line, preferably of instruments with low inharmonicity, such as recorder, flute and many others.
Here is an example to show its capabilities. The original bird song audio recording has been transcribed by Audio Pitch Tracer and can then be played from within Tune Smithy, on your synth or any MIDI instruments. The transcription can be exported as a .csv database or MIDI file to use in any other program.
You need to click on the link to play it. Sorry - there are issues with the embedded player under Firefox so can't embed midi files directly into this page (the embedded player freezes, presumably while pre-processing the midi file, maybe because of all the pitch bends in it as it works fine with ordinary midi files).
3 - the same transcription played on MIDI French Horn, transposed down 2 octaves and slowed down:
Audio Pitch Tracer excels in transcriptions of microtonal nuances of pitch. If you care whether the note was C, or C an eighth note sharp, or C 3.5 cents sharp, then this is the tool for you. If you just want the nearest concert pitch note to the sung pitch, then there are far better ways to do it.
It's other main advantage is that it has extraordinarily good time resolution. If you want to follow the exact pitch of every detail of e.g. a bird song in midi, at a resolution of milliseconds, then again this is the tool for you. I have yet to find any other tool that can transcribe birdsong like this and no-one else has yet told me about any other tool that can do this
So it is useful for:
Microtonal composers who want to make a transcription with the exact pitch in cents of every note played or sung.
Anyone who wants to transcribe birdsong with all its nuances into a midi file so that you can play it on different midi instrument sounds and transform it in ways you can only do in midi
If this is you, then the occasional glitches you may get are a minor nuisance, as the software does something that no other program (to my knowledge) can do at all, and gets you a long way towards your goal even if it isn't quite perfect.
Glitches with this method :
Can find extra notes in background noise. This can be fixed with care - ideally if you can do it, start with a very clean recording - then reduce the noise as much as possible in software such as Goldwave. Low pitched nose such as traffic, water, wind especially - easily done with a high pass filter. Then removing noise and background sound such as distant birdsong is also important, again you can easily do this with noise filters.
Can't cope with any amount of polyphony. Polyphony is surprisingly common in birdsong, some birds like thrushes are renowned for it. Birds have two vocal chords and some species use them independently to sing two pitches at once. When you analyse it in the pitch tracer it will usually pick up the higher pitch. To a human ear if you don't pick up on the polyphony (which can be subtle), it usually sounds more like the lower of the two pitches. This again can be fixed by going through the recording note by note with e.g. Goldwave, and removing either the high or the low pitch component as desired with a low or high pass filter - then retranscribing. But takes a while to do that of course
If you just want to transcribe bird song to twelve equal, and you are a musician and can transcribe melody lines by ear, and don't care about the nuances of pitch for your work, then you can do it easily yourself with other software.
To change the pitch of the birdsong and slow it down at the same time, try GoldWave (probably many other audio editors can do the same thing).
You just load the audio file in GoldWave, then go to Time Warp in the Effect menu in GoldWave, then you can set the amount of the change to 25% to slow down to quarter speed. It's really easy to use. This is shareware software - with a very generous trial policy. Easy to use
If you want to slow it down further, e.g. an eighth, you can repeat the process (time warp by an extra 50%).
Another useful looking tool for musicians, which gets good reviews is: xscribe - it doesn't do the transcription for you. But useful tool to help with transciption by ear.. Slows down audio to help you transcribe it, optionally changing pitch as well, with bookmarks and other features to make it easier to transcribe the audio.
Maybe you could use both.
Apart from that your best bet is the Audio to Midi software. Many tools available. I list a few on the Links page though it is likely to go out of date quickly and not updated for some time now. Try a google for "Wav2midi" or "Wave to Midi".
I haven't needed to use these myself so can't offer an evaluation. Tested a few and seemed like they might be useful for some situations. The ones I tested didn't work 100% but got most of the notes right - so you would need to go through afterwords to fix the notes it got wrong. The field may well have improved since then, was some years ago now that I programmed this software and did the research.
The Audio Pitch Tracer is currently a bit tricky to use
Sorry about this. I did the best I could at the time but have learnt a lot about user interfaces since then. The feedback I get from users is that this feature is tricky to use, and indeed, when I look at it myself with a fresh eye, I can well see your point!
Will do an update of it in version 4.2 when it is ready.
Meanwhile if you need help do contact me and I'll give you some tips to get started and a demo project.
I can also transcribe birdsong for you to midi and provide a transcription similar to the ones on this page at $40 per song (lower rates if unemployed or in education). That's because I'm basically charging at similar rates to a freelance coder for the service, because if not doing this I could be writing my own software instead.
You can do it quickly in the software yourself once you know how so if you have a bit of time to spare to learn how to do it, do ask me how to do it and I'll take you through it step by step. There is no fee for that currently, except that you'll need to buy the software, once only payment, similar to price for transcription of just one song.
Also if you check back in a few months time, I may well have updated it to a much easier to use program. Do ask if you want to be notified when the new version is ready.
Using the Audio Pitch Tracer to transcribe it down - up
This is what the transcription looks like in Tune Smithy's audio pitch tracer:
The red lines here show the start and end positions of all the pitches found
And this is where you would transcribe to MIDI:
The Transcribe to Theremin option, selected in the screen shot, is especially useful for bird song - as it gives those large pitch glides and (optionally) the subtle inflections of volume within every note. It was used for nearly all these transcriptions.
To find this feature in Tune Smithy (FTS) - go to the Audio Pitch Tracer task in your Tune Smithy Tasks folder.
Assuming that everything is already set up for the type of transcription you want to do - then what you do is to open an audio file, and then click the Find & Transcribe to Seed button. I call the found notes a seed, because it can be used as a seed for the fractal tunes in tune smithy. If using the theremin, use the Find & Transcribe to Theremin button - only the seed however can be used with the fractal tunes.
The Play Seed button plays the seed. For the theremin, then you set it to play when you click the Transcribe button, and it plays as it transcribes.
If you want to make a recording, then switch on the Record to Midi on Play, and set the file name to record to using the Record To File button - you can also use the shortcut Ctrl + 11 to bring up this window.
Or, use the Save As Midi File button. That's better for the subtle inflections, particularly the volume fluctuations, as it lets FTS spend as long as is needed to trace the original volume variations in detail - it may spend more time preparing the midi clip than it takes to play it back.
Tip: hold down Shift, Alt, or both of those, while you press the Play or Play Detail button in Tune Smithy's Audio pitch tracer - then you hear the recording at half, quarter, or eighth speed (depending on which of the keys are held down). It is done by just slowing down playback, so it will also shift the pitch down by one, two or three octaves. This is especially useful for hearing the details of high pitched fast sounds such as bird song.
Then you can use the same key combinations with the Play Seed or Transcribe to Theremin buttons to hear the result played in the same way - slowed down and lowered in pitch - to allow easy comparison with the original recording for the rapidly changing notes in the birdsong.
To configure the way the notes are found, you use the Opts for Find Notes window ( Ctrl + 73 ). It has some presets for various types of sound to get you started.
Bird song recordings are often noisy - perhaps natural sounds like wind, sea breakers, or running water - or traffic noise - or more distant bird song higher or lower in pitch. But it is fairly easy to clean them up.
You can do the noise reduction and bandpass with any modern audio editor. Then if the recording still has a fair bit of residual noise in between the notes of the song (e.g. distant bird song) then go through the recording setting it to silence between every burst of song from the bird you want to transcribe. I used all these techniques with the musician wren transcription.
As for the tool to use for this - I use Goldwave and recommend it. For a free solution you can try Audacity. I'm sure there are many other possibilities.
To transcribe birdsong like this, you need to be able to catch even the minutest nuances of pitch. This means the transcription has to be to be as exact as possible - the usual nearest semitone accuracy is nowhere near good enough to get a transcription that sounds like the original. For birdsong, it has to be accurate for notes as short as a hundredth of a second or less.
Normal methods used in this field don't work to quite this level of precision in the pitch domain for notes as short as this. Generally it is considered adequate to find the pitch to the nearest semitone, which is no use if you want to transcribe all the nuances of bird song, or for microtonal transcriptions.
It is possible to transcribe by ear to an adequate level of precision, close enough so that the transcription sounds like the original, but it does take a lot of work. I tested this when I did a short transcription of one phrase of a robin's song by ear for an earlier version of Tune Smithy. It was just a matter of taking short sections of the song, about a hundredth of a second or so, then you play them in a loop at say quarter or eighth speed - then adjust a reference pitch until it is in tune with the song. Doing that for a hundredth of a second doesn't take that long - but as you can imagine, transcribing the whole song takes ages.
You will almost certainly have to do some preparation of your recordings to remove the noise - unless they are taken in completely silent conditions. You may need to do some tweaking of settings in the Audio Pitch Shifter too (especially if there is even a tiny bit of residual noise left). But it certainly beats doing it all by ear :-).
For details of how the Audio Pitch Tracer method works, with a slowed down audio recording for detailed comparison of the midi and the audio, see the How it works page.
Freeware / Shareware status: This feature is shareware.
You are recommended to try out the program first. If you decide to purchase, there are several options available depending on your requirements.
You can get the Midi Relay level to unlock this feature completely.
If you want to use the transcribed seeds to make fractal tunes, get the Complete level.
Occasional use order types
If you use this feature only occasionally or in a minor way, the Midi Save level, will unlock the save to midi files, and also let you transcribe any part of a recording up to one minute long. This may be enough for a typically short burst of bird song for instance.
The Play level lets you transcribe any part of a recording of up to four minutes, and also unlocks the fractal tunes.
The freeware mode for this feature lets you transcribe ten seconds of the song. This is enough to transcribe a short single phrase of bird song - the musician wren phrase on this page is about eight seconds long.
You can upgrade later on if you buy one of the more limited levels.
To continue reading about the Audio Pitch Tracer method, go on to the Bird Song - which gives more detail of the bird song transcriptions.