Hi! I am glad to return after a long period of inactivity on the forum.

During that time I was trying to build a web service, which allows everyone to upload music in MIDI format and get it performed with professional virtual instruments without installing any software: https://artinfuser.com

Most of orchestra instruments are implemented and also piano, pipe organ and jazz drum set are available. You can upload your music and try. I hope that this service can be useful for students and people who do not have virtual instruments installed. And also for composers who do not want to spend time on tweaking virtual instruments. 

You are able to tweak hundreds of parameters for each instrument, for example increase/decrease length of automatic crescendo, decrease use of glissando articulations, shorten legato transitions, increase or decrease vibrato intensity and more.

Here is how it is done:

1. You upload MIDI file, choose setting and start processing

2. Algorithm uses multiple algorithms to select appropriate articulations, randomize parameters and draw curves for dynamics, vibrato intensity, vibrato speed - thus creating "prepared" MIDI file.

3. Prepared MIDI file is imported into DAW with virtual instruments loaded. Then music is rendered into MP3.

4. Now you can download MP3 or prepared MIDI file as you want. You can even download separate MP3 files for each instrument.

You can listen to an example of audio that  is created automatically on the main page: https://artinfuser.com/studio

You need to be a member of Composers' Forum to add comments!

Join Composers' Forum

Email me when people reply –

Replies

  • Thanks for added information. So we are talking about this excerpt:

    http://artportal.su/ctools/job.php?j_id=532

    Here are separate tracks for each instrument:

    http://artportal.su/ctools/stems.php?j_id=532

    - I agree that trill sounds bad. Actually, in this MIDI file trill is exported as 1/30 staccato notes, which is not how the trill will be executed by musician: it will be legato and probably slower. I am planning to detect trills and tremolos and play it more realistically later.

    - I checked note endings, they seem natural to me, frankly speaking I did not find cut offs. Maybe you could point some particular second in particular instrument track for better understanding?

    - Sixteenth-note pattern seems to be played correctly and realistically to me. What do you think is wrong?

  • HI, Alexey,

    Yes, those are the tracks.

    For note endings, I meant the last note before a silence. Look at the waveform in the files I attached and compare them to your waveforms.

    I just listened to the flute sixteenth note pattern at 00:28.032 in both the sample audio file I sent you and the version you produced. Your version sounds like samples that have been glued together while the NotePerformer version sounds like naturalistic playing even though the sampled flute is not as good. You might want to look into using SynFul or, at least, check out their samples and see if you can match them. They have samples on their home page and on their User Demos page.

    I'm no expert, but maybe if you add some resonance, it might smear the sound enough to hide the problems.

  • Hi Alexey,

    I was curious to listen having read the posts and I agree with Antonio about adding ambience. I do also think that Antonios' sounds are closer in timbre to the real thing, as well as the legato performance and overall feel. What sounds are you using for your playback? 

  • Hi! Thanks a lot for detailed answers.

    I use latest versions of sample modeling for all woodwinds and brass.

    Last flute note is short, this is why it sounds strange. You can increase length of note and it will sound ok.

    I uploaded NotePerformer tracks to the same folder with Composer Tools render so that one can conveniently compare each instrument: http://artportal.su/ctools/stems.php?j_id=532

    Please pay attention that instrument tracks of Composer Tools do not contain reverb late reflections, they are added only to master. This is why you hear room reverb lacking in instrument tracks. Actually, listening to instrument without reverb late reflections can be not easy if you are not used to, it can seem to sound artificual. But this allows to hear all articulation problems very distinctly. Moreover, I tend to limit reverb amount to avoid mess when more complex textures are played. I think that I will allow user to control amount of reverb later, I already have ideas how to do it.

    Frankly speaking, I would say that I hear the opposite - that in NotePerformer clarinet track at 0:28 samples are glued together without transition, which sounds artificial, because I do not hear how player changes notes. Anyways I think that difference is not very outstanding here.

    I think, that NotePerformer and SynFul definitely can have their lovers, although they sound to me artificial (lacking realistic sound) when compared to sample modeling, which I use. I actually posted several comparison videos above with NotePerformer and Composer Tools playing same music.

    Yet there seem to be a lot of things to do to use whole potential of sample library in Composer Tools. Currently trills are played bad, also passage starting from 0:59 sounds not very convincing. I will fix that later.

  • Well, the world of sampling is full of disagreements. I've often released a piece and had someone tell me that one instrument sounded fake while another person raved about how real it sounded. One time I posted a piece with an oboe part I didn't think sounded that realistic and then had an oboist ask me "Who was the oboist performing in that recording?"

    If you don't hear what I hear, well, there's not much more to be said. By the way, I mentioned the flute part at 0:28, not the clarinet. To me the difference is like night and day. I included that passage deliberately as I know it's a tough test of legato playing for sample libraries (EWQLSO doesn't do a good job, either).

This reply was deleted.