Neuratron's AudioScore Ultimate 8 seeks to automate a task that few musicians enjoy doing, that of transcribing audio performances of music into written music notation. The technology has come a very long way since its early iterations but is it solid enough to bother with? Read this review to find out!
AudioScore Ultimate 8 from Avid has vastly improved its audio recognition and analysis capabilities as compared to the versions that I tried even a few years ago. Today the product is capable of analyzing polyphonic music scores from a variety of sources and in all honesty does the job fairly well considering how complicated this task is. One of the biggest problems of earlier versions was how percussion sounds would play havoc with the transcription at times but AudioScore 8 improves that significantly by adding new algorhythms that filter out the percussion and make it less of an issue. This along with many other improvements have made the recognition of performed audio much more accurate in some situations and with some types of performance inputs.
Transcribing Live Solo Instruments With AudioScore Ultimate 8
My first test with AudioScore 8 was to try performing some simple songs on solo instruments such as the trumpet and the clarinet. Playing a simple rendition of Hot Cross Buns or Are You Sleeping got me familiar with the interface and how AudioScore works its magic. I noticed right off the bat that the system clearly notices every imperfection that might be present in my playing, including any unintentional scooping or intonation issues with my notes. As far as pitch accuracy in the transcription though I have no complaints. AudioScore always accurately determined the pitch I was playing, and automatically determined what the instrument was that I was playing as well. This came in handy when I told it to transcribe the notes into notation and it automatically transposed the pitches into the correct notes for that instrument rather than showing them in concert pitch.
The only difficulties I had with the transcription was in the realm of note durations (view the video above or here on YouTube). Apparently I like to lead the beat a little bit and as a result the rhythmic parts of the transcriptions were often very messy and required some additional adjustments. That having been said I am a newbie to AudioScore and their help files do go into many different things that you can do to improve the transcription quality. For those with the time to figure these adjustments out I am sure it would make things much more accurate, but out of the box the transcription quality depends greatly on the accuracy and abilities of the performer.
Transcribing MIDI Based Performances With AudioScore Ultimate 8
Transcribing MIDI played in from a keyboard into AudioScore 8 was much less problematic than with the microphone. This is probably to be expected as it is a purely digital representation of the notes, but it bears being said that this is truly the way to go if you have the ability to do so. Transcribing my performances via MIDI was much more accurate and much less confusing. Connecting to my M-Audio Keystation 49e keyboard was quick and painless and recording was equally so. My only problem came in that, as with the live acoustic instruments, I tend to lead the beat just a tad, as a result the first note of my recordings almost always was missing because the MIDI on keypress came barely before the system actually started recording.
In MIDI mode however AudioScore Ultimate 8 shined not only with pitch accuracy but also with rhythmic accuracy. No problems there both with single lines of music and with polyphonic performances. Once the music was in the system I can quickly export it as standard music notation in MusicXML format and import it into music notation programs like Sibelius or Finale.
Transcribing MP3 Recordings Into Music Notation With AudioScore Ultimate 8
One of the most intriguing parts of the AudioScore transcription software is that it can also transcribe scores that originate as MP3's and other recordings of music. To test this out I had it sample a variety of different recordings. Some very simple melodic pieces played by single instruments and others with more complicated polyphonic sources. Single line melodies were fairly easy for AudioScore to transcribe and for the most part it did a good job of pitch accuracy although fine tuning the rhythms still required additional tweaking.
Anything with multiple instruments makes things much more complicated and thus much less accurate. I had the system try to transcribe a duet recording of string instruments and while it did catch all the notes the transcription of those notes was all but unusable. Again, with more time and more dedication to the task I would have been able to single out the specific instruments, break them out into parts, and clean up the transcription significantly. Remember though that even humans have problems performing this kind of multi-instrument transcription task so this alone is not a major downfall of the program.
When used in the right conditions and with the proper mindset going into it AudioScore 8 does it's job as a fast, effective way to transcribe audio music performances into notation. It has it's limits, but those limits seem to decrease with each iteration of the program. As mentioned before when AudioScore is presented with a clean, tonally accurate input performance it can make a challenging, difficult task into one that is much less stressful.