RapidComposer v5 beta feedback and discussion

[magicmusic]

Is it also possible that you can interpret bend instructions chat gpd give ?. bends makes all in general sound much more exciting. see the video, he use it with chat gpt and chat gpt output this instructions at 6:30

Hi magicmusic,

yes, this is possible. I managed to get bending instructions with the notes. But how do you want to include the bends in your song? Like inserting an articulation that your guitar synth will interpret, or adding pitch bend messages? Another idea is to have 'bend up' and 'bend down' presets stored in the CC browser that could be added to bent notes.

Thanks,
Attila

I think such bend up bend down presets in articulation browser the best. RC did then add this in the controller track of the phrase. this can also done with vibrato presets .bend on guitar is same as on keyboard +2 semitones possible

for slides that go over more than 2 tones RC can create a little overlapping of the next note. there can choose or create keyboard sounds with portamento do then the slide over more than 2 notes

[BluGenes]

Need a way to copy and paste the chat posts.. so you won't have to type everything again situation..

[BluGenes]

Actually, for the prompts, you should follow the idea, like the image AI stuff follows for prompts. Have a way to store favorite prompts to reuse. (kind of like automatic1111 does)

Also, the prompt entry area needs to be wider to see more text for catching typos.

I also think you should try to use Studio LM as a alternative for the AI, or, just another choice for AI, like we discussed before.
It would give the user the ability to have everything installed locally. It does use the LAMA model, which does seem to best for music generation. The only thing that I think that is a problem with it is the size of it, depending on which one you install, that is. You can can also use the same calls to it from CURL, so, nothing changes there.

[lovemusic15]

If you skip to the 56-second mark in this video, you'll see precisely what I've been discussing. Imagine applying this idea to RC, leveraging its capabilities for MIDI. While the video focuses on audio, the same concept can be immensely powerful when adapted to MIDI. Picture inputting your preferred melody and utilizing the Rhythm processing feature with the enhancements I previously mentioned. With these options, you can input your favorite melody and use the Rhythm processing to generate music. The fusion of these features not only aligns with what I've requested earlier but also aligns with the broader vision of AI, showcasing what users truly desire in software. This insight should provide a clear understanding of the potential enhancements that could be made


it's nice to be able to just put prompts of what you want but even better when you can have a MIDI phrase as a starting point

[musicdevelopments]

think you need a way to clear the edits for the AI or some sort of reset. Once you reach the limit, there isn't a way to fix the issue..

Actually you can fix it. You can start a new conversation by pressing the 'trash' button:



You rarely need as many input tokens, only if your next question refers to past questions and replies. If you ask a new question, it is a waste of tokens (and money) to send all previous questions and replies.

Thanks,
Attila

[musicdevelopments]

The structure of a simple 'pop/rock/etc' song is asked for, so I guess there are not many different replies. I included 'simple' because the user can add verses if needed. You can try increasing the temperature to get other replies.

[musicdevelopments]

Need a way to copy and paste the chat posts.. so you won't have to type everything again situation..

Good idea!

[sj1]

The structure of a simple 'pop/rock/etc' song is asked for, so I guess there are not many different replies. I included 'simple' because the user can add verses if needed. You can try increasing the temperature to get other replies.

Hmmm. I understand you can't achieve more here with the AI than what the AI actually offers. In my own experiments, wrt. a large variety of song structures, what's on offer is pretty weak. Maybe at some future point that will change.

Still though, I would like to commend to you Attila the series of posts starting here where I comment at some length on a way RC could deal with structure generation and accumulation entirely on it's own, without a major paradigm change.

viewtopic.php?t=600601&start=215

[plopseaw]

Tags and Apple Silicon Compatibly

First, Ive got a big midi library now. I'm using custom tags but I'am wondering how I can search for two tags at the same time. For instance searching for house and stabs. Should it be {house)!="" && (stab)="". Doesn't seem to work. As the library gets bigger the need to tag and search greatly increases.

Also I notice on (Apple Silicon, Bitwig, Sonoma, M2 Mac book pro) the speed on entering text into the search field is really slow (like one character every 2 seconds)

In general the apple silicon version V5 is very hit and miss. if I enable detect retina display Then I find the font doubles in size every time I open the vst. I disable it and also using metal for ogengl as with meta the screen is very fuzzy.

Also on Apple silicon, the vst freezes the display. I need to close the vst and open it again for the playhead to move or to be able to use the scroll bars etc. Love the program but very Janky on apple silicon.

[lulukom]

Merry Christmas everyone!
I don't really understand this, but I'm wondering: is the RC connected to one neural network or is it possible to use others? Thank you .

[Jmf1928]

Are there any known issues with Bitwig 5 on PC? I have no problem getting MIDI out into Ableton as a host, but with Bitwig I'm not getting any MIDI out of the plugin version.

[plopseaw]

I'am using note receive in Bitwig to move midi around (on a Mac but should be the same as PC). I put RC on a track on its self and then use note receive set to RapidComposer and the midi channel of the track (1,2 etc). After the not receive I put a channel filter (1,2, 3 etc). works very well

[BluGenes]

hey Atilla, was thinking and have a question on the AI..

First, maybe some sort of library for prompts, so we can re-use them.

The question is, is it possible to set up the AI (on your side) so that we can create polyphonic phrases? I think it would be cool to be able to have a prompt for something like that, as an example, "create an arpeggio using both the 3rd and 7ths on the 1st and 3rd beats and any scale note on the 2nd and fourth beat and a one random rest on any of the off beats" , or what ever..

currently it seems it only creates monophonic phrases..

[BluGenes]

also forgot to mention, any of the GPT's that have "instruct" in them are actually only for assistants, like co-pilot for VS. Those are designed for coding style help.. so, maybe drop those from the list..

[BluGenes]

yea, just realized my example request above isn't an arpeggio, lol, but, still, can't get anything polyphonic