Audio to MIDI converter

[joecass]

Help ! I need a solid program that will convert audio files (either .wav or MP3) to MIDI. I've tried several programs that I downloaded, none of them worked correctly. If anyone knows of such a program (preferably free), I'd appreciate any advice. Just want to convert a simple stereo piano .wav file to an accurate MIDI file that can be played through a synthesizer.

[sum_guy]

I've never had any success doing wav/mp3 conversions to midi, if you have 1 single instrument in your wav/mp3 file and nothing else, you may get a result that's not quite cats screaming. There are some tutorials online using Audacity and others, you probably won't be happy with any of the freeware solutions, and the ones that do work pretty good are really expensive.

[Cornucopia]

This kind of process is a complex pattern recognition challenge for even the best algorithms. For real-world, multi-instrument and/or noise-infused sounds, it's basically not feasible at this time.
If your "simple stereo piano.wav" file is not polyphonic (only one note at a time = no chords), you should be able to find stuff that can do it. Polyphonic material can be doable as long as it isn't too rich in harmonics.
(Personally, I wouldn't even CONSIDER using mp3 as a source file: the algorithm in use by MP3 does its work by throwing out tones - if those tones are note fundamentals, you are losing exactly the info you need to generate a proper transcription.)

All the programs that I know of that do this well are NOT cheap, let alone free (usually built into or as plugins for DAW or Music Composition/Notation/Scoring software, such as Sibelius, Finale, ProTools, Reaper, Logic, etc).

For something a little more appropriate & affordable, I'd suggest something like IntelliScore.

*Note that with ANY/EVERY of these programs, just as with OCR or Voice-to-Text, you will have to do your own manual cleanup/adjustment to get it close to what you are intending.

Personally, I've found them not that helpful overall, and I usually just transcribe manually, by ear (occasionally using one of those tools to compare against what I've done). Human analysis is invariably better, ATM.

Scott

[joecass]

This kind of process is a complex pattern recognition challenge for even the best algorithms. For real-world, multi-instrument and/or noise-infused sounds, it's basically not feasible at this time.
If your "simple stereo piano.wav" file is not polyphonic (only one note at a time = no chords), you should be able to find stuff that can do it. Polyphonic material can be doable as long as it isn't too rich in harmonics.
(Personally, I wouldn't even CONSIDER using mp3 as a source file: the algorithm in use by MP3 does its work by throwing out tones - if those tones are note fundamentals, you are losing exactly the info you need to generate a proper transcription.)

All the programs that I know of that do this well are NOT cheap, let alone free (usually built into or as plugins for DAW or Music Composition/Notation/Scoring software, such as Sibelius, Finale, ProTools, Reaper, Logic, etc).

For something a little more appropriate & affordable, I'd suggest something like IntelliScore.

*Note that with ANY/EVERY of these programs, just as with OCR or Voice-to-Text, you will have to do your own manual cleanup/adjustment to get it close to what you are intending.

Personally, I've found them not that helpful overall, and I usually just transcribe manually, by ear (occasionally using one of those tools to compare against what I've done). Human analysis is invariably better, ATM.

Scott

Yes, it's a simple polyphonic piano track, as opposed to a full-blown song containing many instruments. I tried Intelliscore and other programs, none of them come close. They even claim to be able to convert a song containing drums, bass, guitar, vocal, etc. to MIDI, which seems impossible. I don't use DAW software or plug-ins..... I do my own home recordings myself, except I can't play piano. BUT.... if I can get this piano track converted to MIDI, I can correct the timing and edit the parts individually that I need, convert it back to an audio file, and insert it into my recording. I tried using three different programs using both .wav and MP3, couldn't get anything doable.

[netmask56]

I would say it is a thankless task and you end up doing as cornucopia described. Another trick is to make an audio track, preferably wav and slow it down without altering pitch - this might help the recognition program to at least get the right notes but you will still have a fair amount of editing in your sequencer program.

[tugshank]

cubase user here... you need a human being to transcribe/track that file for you <cracks knuckles>.