Can music be regenerated virtually?
-
Let me explain...A lot of people are trying to find compression techniques for audio.... I just had a thought if we had a framework like "Music player framework (MPF)" that would have all the musical instruments virtualized. Well that's very much possible . Example Microsoft's Direct Music producer Most songs/music have repetitive sounds...like a drum beat every 2s and a flute played in background... If we can analyze the WAV audio and therby create a music-player-program(MPP) that would run on MPF, wouldnt that lead to a lot of compression? Vocals cannot be programmed until machines can sing songs, but vocals take only a fraction of memory in the audio file Does anybody know any such kind of work done before? Anyways, today, I did invent two new terms MPF and MPP :laugh:
Anup Shinde wrote:
Most songs/music have repetitive sounds...
I take it you don't play an instrument yourself? Although there is repetition to a degree you can't fake the subtle inflections of a passionate guitar solo. Long live real music I say. And no, dance music is not real...
-
Thanks guys, i got your point....and now my inventions proved waste :(( But I still have question(s) :) MIDI is small size and even plain vocals are small size (they dont need much bandwidth)....then why are the MP3 so big size files (5 MB is big enough) ;P cud be silly question.....but i like asking stupid questions :laugh:
[oversimplification mode ON] There are two ways of representing audio in a computer: as a waveform (PCM and many more) and as a sequence of instructions (MIDI). Waveforms are accurate, they just note down thousands of times per second the value of the audio wave, and any usual speaker can be used to reproduce them. A single song in wave form easily takes 50 MBs of space. MP3s are an evolution of this technique, with an amazing, lossy, compression algorithm, but they still have to contain the values of the sound wave. There are algorithms with better compression ratio, MP3 is just the most popular. MIDIs on contrary just keep track of the "sheets": notes, instruments, time. Sure they're smaller, but then they delegate the rendering of the audio wave to the listener device. This means that MIDIs do not sound the same across different PCs (try with a pc equipped with different audio cards and you'll hear it by yourself). It's the same difference between a 50$ keyboard and a 30k$ Grand Piano: they do not sound the same, while they both have keys. They're consequently not suitable for hi-fidelity audio distribution. Moreover, it is not possible to create a midi from a waveform (while the contrary is easy), since waveforms are a single signal composed by the sum of vocals and different instruments. And, of course, there's the issue of post-processing filters: many songs we have on CDs are heavily manipulated after recording with various techniques. These filters are not easy to represent with a MIDI notation, while in waveforms we hear their result.
Luca The Price of Freedom is Eternal Vigilance.
-
Thanks guys, i got your point....and now my inventions proved waste :(( But I still have question(s) :) MIDI is small size and even plain vocals are small size (they dont need much bandwidth)....then why are the MP3 so big size files (5 MB is big enough) ;P cud be silly question.....but i like asking stupid questions :laugh:
Because an mp3 is real music, and a MIDI file is information that a MIDI synthesizer uses to imitate real music.
Christian Graus - Microsoft MVP - C++ Metal Musings - Rex and my new metal blog "I am working on a project that will convert a FORTRAN code to corresponding C++ code.I am not aware of FORTRAN syntax" ( spotted in the C++/CLI forum )
-
[oversimplification mode ON] There are two ways of representing audio in a computer: as a waveform (PCM and many more) and as a sequence of instructions (MIDI). Waveforms are accurate, they just note down thousands of times per second the value of the audio wave, and any usual speaker can be used to reproduce them. A single song in wave form easily takes 50 MBs of space. MP3s are an evolution of this technique, with an amazing, lossy, compression algorithm, but they still have to contain the values of the sound wave. There are algorithms with better compression ratio, MP3 is just the most popular. MIDIs on contrary just keep track of the "sheets": notes, instruments, time. Sure they're smaller, but then they delegate the rendering of the audio wave to the listener device. This means that MIDIs do not sound the same across different PCs (try with a pc equipped with different audio cards and you'll hear it by yourself). It's the same difference between a 50$ keyboard and a 30k$ Grand Piano: they do not sound the same, while they both have keys. They're consequently not suitable for hi-fidelity audio distribution. Moreover, it is not possible to create a midi from a waveform (while the contrary is easy), since waveforms are a single signal composed by the sum of vocals and different instruments. And, of course, there's the issue of post-processing filters: many songs we have on CDs are heavily manipulated after recording with various techniques. These filters are not easy to represent with a MIDI notation, while in waveforms we hear their result.
Luca The Price of Freedom is Eternal Vigilance.
-
Thanks for the real detailed info. You said that MIDIs donot sound same across diffrent PCs. But WAV audio will sound the same everywhere. Then if I go thru the path MIDI->Rendered WAV->Soundcard, it should be fine ...right?
I'm not sure about what you mean. If you mean "if I copy across computers the rendered WAV, it will sound right", you're right, but you'll be trasferring a 50MB WAV, not a MIDI. Remember that the transition MIDI->Speaker is usually done by the soundcard, in hardware. A software MIDI rendering is possible of course, but home-synthetized instruments will never sound like real instruments, there's too much detail in the acoustic field that is not encoded by MIDI. Can you describe using your own words the difference between a keyboard and a piano? Imagine doing that in mathematical terms...
Luca The Price of Freedom is Eternal Vigilance.
-
Let me explain...A lot of people are trying to find compression techniques for audio.... I just had a thought if we had a framework like "Music player framework (MPF)" that would have all the musical instruments virtualized. Well that's very much possible . Example Microsoft's Direct Music producer Most songs/music have repetitive sounds...like a drum beat every 2s and a flute played in background... If we can analyze the WAV audio and therby create a music-player-program(MPP) that would run on MPF, wouldnt that lead to a lot of compression? Vocals cannot be programmed until machines can sing songs, but vocals take only a fraction of memory in the audio file Does anybody know any such kind of work done before? Anyways, today, I did invent two new terms MPF and MPP :laugh:
Anup Shinde wrote:
If we can analyze the WAV audio and therby create a music-player-program(MPP) that would run on MPF, wouldnt that lead to a lot of compression?
Yes, and a lot of loss.
-
Hrm.. I wonder if I can find one of the billion versions of Axel Foley. :)
-- My disbelief is not a belief.
Hahaha, memories!
-
Old idea called 'mod' :). Was the best way of doing music back in the days when Amigas walked the earth. You can get some very nice software http://www.madtracker.org/[^] to make music with this technique. I have heard some pretty decent things done with it, though would be a while before you could match Squarepusher.
-
[oversimplification mode ON] There are two ways of representing audio in a computer: as a waveform (PCM and many more) and as a sequence of instructions (MIDI). Waveforms are accurate, they just note down thousands of times per second the value of the audio wave, and any usual speaker can be used to reproduce them. A single song in wave form easily takes 50 MBs of space. MP3s are an evolution of this technique, with an amazing, lossy, compression algorithm, but they still have to contain the values of the sound wave. There are algorithms with better compression ratio, MP3 is just the most popular. MIDIs on contrary just keep track of the "sheets": notes, instruments, time. Sure they're smaller, but then they delegate the rendering of the audio wave to the listener device. This means that MIDIs do not sound the same across different PCs (try with a pc equipped with different audio cards and you'll hear it by yourself). It's the same difference between a 50$ keyboard and a 30k$ Grand Piano: they do not sound the same, while they both have keys. They're consequently not suitable for hi-fidelity audio distribution. Moreover, it is not possible to create a midi from a waveform (while the contrary is easy), since waveforms are a single signal composed by the sum of vocals and different instruments. And, of course, there's the issue of post-processing filters: many songs we have on CDs are heavily manipulated after recording with various techniques. These filters are not easy to represent with a MIDI notation, while in waveforms we hear their result.
Luca The Price of Freedom is Eternal Vigilance.
Two things to note about mp3 compression is that it is lossy and use is made of perceptual weighting to determine what to discard. Perceptual weighting means that it tries to find information in the signal that the listener won't be able to hear, and discard it, and because it discards information it is lossy compression. An example of how it works, if during one time segment there is a lot of signal around say 1kHz, relatively little around 3kHz, depending on the weighting rules, most or all of the information of the 3kHz signal will be discarded. This was the result of many listening tests to determine what listeners could distinguish.
Peter "Until the invention of the computer, the machine gun was the device that enabled humans to make the most mistakes in the smallest amount of time."
-
I'm not sure about what you mean. If you mean "if I copy across computers the rendered WAV, it will sound right", you're right, but you'll be trasferring a 50MB WAV, not a MIDI. Remember that the transition MIDI->Speaker is usually done by the soundcard, in hardware. A software MIDI rendering is possible of course, but home-synthetized instruments will never sound like real instruments, there's too much detail in the acoustic field that is not encoded by MIDI. Can you describe using your own words the difference between a keyboard and a piano? Imagine doing that in mathematical terms...
Luca The Price of Freedom is Eternal Vigilance.
Well you got it right....i was referring to software MIDI rendering... The next line that you say: "MIDI cannot encode everything". I think this takes me back to the main question "Can music be generated virtually"" without loss of the ascoutic field that you are referring to?" I was referring to MIDI as some ppl here claim that MIDI can encode every instrument. (I dont have much knowledge to argue with them in this case) Also real music has noise associated with it.. Say when someone plays guitar...there is also a squeaky sound that you can hear....thats noise and has nothing to do with the actual tunes being played... Synthesized music can eliminate that.... and for the ascoutic range I think that every musical part can be represented as a mathematical equation using sines, cosines, etc...prob is that its very complex...
-
Because an mp3 is real music, and a MIDI file is information that a MIDI synthesizer uses to imitate real music.
Christian Graus - Microsoft MVP - C++ Metal Musings - Rex and my new metal blog "I am working on a project that will convert a FORTRAN code to corresponding C++ code.I am not aware of FORTRAN syntax" ( spotted in the C++/CLI forum )
I had to go to your profile to see it was a suicase in Bob's hand...paw...graspy thing and not a glass of :beer:
-
[oversimplification mode ON] There are two ways of representing audio in a computer: as a waveform (PCM and many more) and as a sequence of instructions (MIDI). Waveforms are accurate, they just note down thousands of times per second the value of the audio wave, and any usual speaker can be used to reproduce them. A single song in wave form easily takes 50 MBs of space. MP3s are an evolution of this technique, with an amazing, lossy, compression algorithm, but they still have to contain the values of the sound wave. There are algorithms with better compression ratio, MP3 is just the most popular. MIDIs on contrary just keep track of the "sheets": notes, instruments, time. Sure they're smaller, but then they delegate the rendering of the audio wave to the listener device. This means that MIDIs do not sound the same across different PCs (try with a pc equipped with different audio cards and you'll hear it by yourself). It's the same difference between a 50$ keyboard and a 30k$ Grand Piano: they do not sound the same, while they both have keys. They're consequently not suitable for hi-fidelity audio distribution. Moreover, it is not possible to create a midi from a waveform (while the contrary is easy), since waveforms are a single signal composed by the sum of vocals and different instruments. And, of course, there's the issue of post-processing filters: many songs we have on CDs are heavily manipulated after recording with various techniques. These filters are not easy to represent with a MIDI notation, while in waveforms we hear their result.
Luca The Price of Freedom is Eternal Vigilance.
Luca Leonardo Scorcia wrote:
MIDIs on contrary just keep track of the "sheets": notes, instruments, time. Sure they're smaller, but then they delegate the rendering of the audio wave to the listener device. This means that MIDIs do not sound the same across different PCs (try with a pc equipped with different audio cards and you'll hear it by yourself). It's the same difference between a 50$ keyboard and a 30k$ Grand Piano: they do not sound the same, while they both have keys. They're consequently not suitable for hi-fidelity audio distribution.
Puts on curmudgeon hat. :laugh: Back in the old days when games came of floppy disks and your PC ran on steam generated by burning whale oil MIDI was fairly common for games since wav format music would take up too much diskspace. As a result the quality of the MIDI tables on your soundcard was almost as important as the rest of its audio generation capability, and if you were a serious gamer/musician you had a $500+ midi daughter card thats sole purpose was to give you a really good set of sample tables.
-- CleaKO The sad part about this instance is that none of the users ever said anything [about the problem]. Pete O`Hanlon Doesn't that just tell you everything you need to know about users?
-
Well you got it right....i was referring to software MIDI rendering... The next line that you say: "MIDI cannot encode everything". I think this takes me back to the main question "Can music be generated virtually"" without loss of the ascoutic field that you are referring to?" I was referring to MIDI as some ppl here claim that MIDI can encode every instrument. (I dont have much knowledge to argue with them in this case) Also real music has noise associated with it.. Say when someone plays guitar...there is also a squeaky sound that you can hear....thats noise and has nothing to do with the actual tunes being played... Synthesized music can eliminate that.... and for the ascoutic range I think that every musical part can be represented as a mathematical equation using sines, cosines, etc...prob is that its very complex...
Anup Shinde wrote:
Synthesized music can eliminate that.... and for the ascoutic range I think that every musical part can be represented as a mathematical equation using sines, cosines, etc...prob is that its very complex...
This is part of why MIDI's sound so unrealistic. Real instruments don't just play tones; there are alot of subharmonics both in the instruments and in surrounding objects (the cavity of a guitar, for example, significantly changes the character of the sound). Interestingly, introducing even a small amount of noise into pure signals causes them to sound much more realistic to the human ear, even if the noise has no coherence (what is called Gaussian noise, or white noise). Of course, to get the most realistic reproduction, you'd want noise that includes subharmonics of your tones, so you wouldn't want pure white noise.
------------ Cheers, Patrick
-
Luca Leonardo Scorcia wrote:
MIDIs on contrary just keep track of the "sheets": notes, instruments, time. Sure they're smaller, but then they delegate the rendering of the audio wave to the listener device. This means that MIDIs do not sound the same across different PCs (try with a pc equipped with different audio cards and you'll hear it by yourself). It's the same difference between a 50$ keyboard and a 30k$ Grand Piano: they do not sound the same, while they both have keys. They're consequently not suitable for hi-fidelity audio distribution.
Puts on curmudgeon hat. :laugh: Back in the old days when games came of floppy disks and your PC ran on steam generated by burning whale oil MIDI was fairly common for games since wav format music would take up too much diskspace. As a result the quality of the MIDI tables on your soundcard was almost as important as the rest of its audio generation capability, and if you were a serious gamer/musician you had a $500+ midi daughter card thats sole purpose was to give you a really good set of sample tables.
-- CleaKO The sad part about this instance is that none of the users ever said anything [about the problem]. Pete O`Hanlon Doesn't that just tell you everything you need to know about users?