RapidComposer v5 beta feedback and discussion

[BluGenes]

Yea, that is rather large. As an example, mine is set to 256 samples and I probably could go as low as 128, but, that is because I have really good audio hardware. For me, I experience what you are seeing when mine is above 1024.

The point being, that number needs to be as low as possible for best performance..

EDIT: Also, to change that, you do that from either your audio hardware, or the driver software for it..

oops, I had to look at your screen shot again, yea, if you can, you need to use ASIO for best results.. not MME. MME can work, but, that's it.. won't perform well..

[BluGenes]

This is what mine looks like.. now, bear in mind, since I make videos, my audio through put is kind of layered, well mixed. That is why you see voicemeetter, which is just software, selected.

The audio hardware is an Apollo Twin, but, could be what ever interface you have, and that has an ASIO driver, which manages the buffer size. That is just passed on to the rest of the software,
Hardware ASIO->VoiceMeeter->RC -> speakers, kind of path, if you will.

[BluGenes]

With the release of OpenAI new model, gpt-4o, I can say this works very well in RC now. It's super fast too.

[sj1]

Thanks very much Gene.

I don't know how that MME got in there, but going back to ASIO and 5 ms worth of buffer fixed things up nicely!

[BluGenes]

Hey Atilla.. seems you may not have to wait on GPT-5.. hehe, yea, the new GPT-4o is pretty nice. What I suggest is to look over how you have your rules set up and look into refining them. Don't lose what you have already! lol.., but, I think it could be possible to lower token counts in the rules, you just have to look into it.

The response time is really well now. This is a video I have from my dev stuff on this, well, an update to one earlier. I did fix the memory issue for a composition, but, that isn't shown in this video. ( I keep a video log of this dev project, mostly for me.. hehe)

One thing I finally come to realize, and it is a duh moment for me, lol, is that AI is really really REALLY good with numeric patterns. So, generating rules that define the relationship, well, is now obviously important.

I think it's possible to have the AI "hear" just by defining the frequencies and working with patterns and musical theory in that sense. I have no clue how it is actually done by other developers and such, just an idea I came upon thinking about this.. It's kind of domino effect really. By defining that, then from that you can define the genre and styles.. and so forth..

anyways.. here's the video..
https://youtu.be/t6rvU1ZpZOo

[lovemusic15]

Attila, I'm getting this problem when using two instances of midi Mutator {VST}

[frimpiso]

Hey Attila,

I would also love to receive a link to the latest beta with the full relative note coloring feature.

Question: Is it possible to also distinguish the note labelling between b3 and 3 or b7 and 7?

Currently it is all the same e.g. for a min7 and maj7 chord the third and 7th are all labelled "3" and "7" instead of "b3" and "b7" for the min7 chord.

Thanks

Jens

P.S.: I am currently taking a jazz piano jump start course from Open Studio and Rapid Composer is a blast to record everything I learn.

[BluGenes]

If you could.. in the strum generator, could you add the ability to lock the strum direction so that when something random is generated the strum direction stays the same? Example would be if you would always wanted the direction to be always up and down.. always..

Be sure to fix that bug too.. in the rhythm..

(If you set the rhythm to manual editing it only generates 1024 length notes no matter the grid settings) (when you double click to add a note)

[BluGenes]

Also, any way you can make the new merge midi import to merge at the play head cursor?