1

Topic: AI and music

And this your AI can solve the task of recognition of music? Here for example a composition executed by a symphonic orchestra from 100500 tools to spread out to scores? Or on record to find the executor, a composition title?

2

Re: AI and music

Hello, Stanislaw K, you wrote: SK> Or on record to find the executor, a composition title?

3

Re: AI and music

Hello, Amon_RA, you wrote: SK>> Or on record to find the executor, a composition title? A_R> I some times tried Shazam (for a long time already). If a composition not pop popular at present result any, to put it mildly. Percent 15 it guessed.

4

Re: AI and music

Hello, Stanislaw K, you wrote: SK> I some times tried (for a long time already). If a composition not pop popular at present result any, to put it mildly. Percent 15 it guessed. Well so it . At them the basis permanently extends.

5

Re: AI and music

Hello, Stanislaw K, you wrote: SK> Here for example a composition executed by a symphonic orchestra from 100500 tools to spread out to scores? Sharing 2 - speakers looks approximately so: http://files.rsdn.org/100404/mix.wav http://files.rsdn.org/100404/spk1.wav http://files.rsdn.org/100404/spk2.wav

6

Re: AI and music

Hello, AlexDoberman, you wrote: SK>> Here for example a composition executed by a symphonic orchestra from 100500 tools to spread out to scores? AD> sharing 2 - speakers looks approximately so: . And batches of ten violins spreads out? Notes writes down?

7

Re: AI and music

Hello, Stanislaw K, you wrote: SK> and batches of ten violins spreads out? Notes writes down? Frenchmen couple of years made a quite good piece back, but there they divided different types of tools. (Wind instruments, shock, speech...) Tools it was possible . As a whole it is quite good on demos looked http://bass-db.gforge.inria.fr/fasst/ I Think what to train  to divide batches of ten violins it is possible. (If on frequencies they differ), but to take pleasure solo sounding of each violin separately hardly it will be possible. Strong enough distortions.

8

Re: AI and music

Hello, Amon_RA, you wrote: A_R> Hello, Stanislaw K, you wrote: SK>> Or on record to find the executor, a composition title? A_R> Shazam Not. , MusiXmatch and analogs are able to analyze stupidly a spectrum of a total sound on the basis, and anything else. It works for a situation "something broadcasts from the TV set in a noisy bar", but does not work, if the same product, but execution another. I checked, in particular, on "Night snipers": they have a heap of concert records with variations of the same song - studio identifies, concert - does not take. The person would identify identity of product at all without knowing language.

9

Re: AI and music

Hello, Stanislaw K, you wrote: SK> Here for example a composition executed by a symphonic orchestra from 100500 tools to spread out to scores? SK> or on record to find the executor, a composition title? And you can, with natural intelligence?

10

Re: AI and music

Hello, Stanislaw K, you wrote: SK> And this your AI can solve the task of recognition of music? Can. SK> here for example a composition executed by a symphonic orchestra from 100500 tools to spread out to scores? Not a problem. I saw on the sly DeepSpeech for Russian and for interest did a network, which of pure PCM (without everyone FFT)  a human voice or singing of birds. I think that on a violin to train not more difficult. SK> or on record to find the executor, a composition title? More difficult. It is possible to train  on recognition of a restricted dial-up of tunes, but on each new song then it it is necessary . It is necessary " with storage" (hard/soft attention) is now a first line in researches on AI.

11

Re: AI and music

Hello, , you wrote: SK>> Here for example a composition executed by a symphonic orchestra from 100500 tools to spread out to scores? > when I learned to jingle on the piano any service of selection of chords tried and have been plunged into shock. It really worked and assorted the loaded roller on accompaniment and a tune for two hands, for each tool and a symphonic orchestra?

12

Re: AI and music

Hello, marcopolo, you wrote: SK>> Here for example a composition executed by a symphonic orchestra from 100500 tools to spread out to scores? SK>> or on record to find the executor, a composition title? M> and you can, with natural intelligence? I can, only it occupies a lot of time from me. AI can (to us promise)  processes and has  volume high-speed (on comparing with me / the average person) storages.

13

Re: AI and music

Hello, Cyberax, you wrote:> I saw a C on the sly DeepSpeech for Russian and for interest did a network, which of pure PCM (without everyone FFT)  a human voice or singing of birds. I think that on a violin to train not more difficult.  a human voice too it is interesting, if on an output there will be a shorthand report/subtitles, with quality it is better than at Google in . SK>> Or on record to find the executor, a composition title? The C> is more difficult. It is possible to train  on recognition of a restricted dial-up of tunes, but on each new song then it it is necessary . A C> It is necessary " with storage" (hard/soft attention) is now a first line in researches on AI. We wait we wait.

14

Re: AI and music

Hello, Stanislaw K, you wrote:>> I saw a C on the sly DeepSpeech for Russian and for interest did a network, which of pure PCM (without everyone FFT)  a human voice or singing of birds. I think that on a violin to train not more difficult. SK> Vytsepljat a human voice too is interesting, if on an output there will be a shorthand report/subtitles, with quality it is better than at Google in . Google for some reason does not use the best  for subtitles. Leaking state-of-the-art is about human level of recognition on standard texts. Generally, DeepSpeech it is easily possible to deliver and be played with it. To all I recommend.

15

Re: AI and music

Hello, AlexDoberman, you wrote: AD> I Think what to train  to divide batches of ten violins it is possible. (If on frequencies they differ), but to take pleasure solo sounding of each violin separately hardly it will be possible. Strong enough distortions. Not the expert on game on a violin, but in the same place can sound simultaneously some strings? How to understand, the sound of what string concerns what batch?