Hey JT, cool to have you. In my perception you[1] and @ycollet[2] have been kind of keeping the lights on as far as “more than single packages” go, very appreciated!
as the maintainer of the audinux COPR ↩︎
Hey JT, cool to have you. In my perception you[1] and @ycollet[2] have been kind of keeping the lights on as far as “more than single packages” go, very appreciated!
as the maintainer of the audinux COPR ↩︎
The past weekend’s OpenAI turmoil has me thinking (about many things! - but one in particular >) that at some point AI will have to be a component in any credible / useful audio tool set. And it’s likely going to be the fastest moving target. And probably the most contentious. And how does open source keep up? If an active SIG is formed here then I do believe it will have to grapple with that. Probably sooner rather than later.
I could be wrong but I dont think there are many open source AI-music projects right now. I know there have been a few projects in the past done as University projects, but IDK if they ever got fully open sourced.
Most of the open source AI focus seems to be around text-gen and image-gen. I havent looked into AI music creation much yet, I wonder what kind of hardware it requires and if its within the reach of most common systems.
It’s only one tool, but Spleeter (which was developer by the streaming service Deezer) is open source. It’s a TensorFlow-based tool which lets you split music into different stems (say vocals, piano, bass, drums & other). The individual stems sound like hot garbage occasionally – something happening in one stem causing artifacts in another – but it’s very useful for practicing songs or transcribing.
Let’s cross that bridge when we get to it? I don’t see the SIG as a major driver behind progress in open source audio software in general, at least not yet . I think our niche is integrating already existing things [1], documenting what has to be set up manually, rounding off the edges and eventually applying polish. I mean, if folks developed, say, an open source plugin to generate drum accompaniments matching someone’s guitar performance using machine learning, they would of course be welcome here, but its development would still be out of the scope of the SIG. Let’s keep focused and not spread ourselves too thin.
Without a formal decision process, let me try to sum up the consensus (filling in some blanks):
This is your opportunity to lodge objections!
With that said, I’ll probably put some work into the Wiki pages and/or find out who can help us with the Matrix channel and Discourse tag next. Assuming the above will stand, if not we can make changes later.
I can help with the Matrix channel. Just need who will step up as admin and moderators.
Oh, I in no way meant to suggest the SIG doing development or leading in AI for audio. Just meant it as keeping a watch out for this and making informed choices as we go. (Plus there was a bit of ‘thinking out loud’ ion my part.)
I just realized I don’t have permissions to create a room under Fedora space. Could anyone in this thread or commops-team admin assist?
I will open a ticket to add JUCE to fedora repo and will put you as a reviewer.
Thanks a lot for the sponsoring
At least during work days I have the Discourse window open, so I’m happy to cover that time.
If you just want a ‘unofficial’ room, you can just make it have a
:fedora.im canonoical alias. If you want a :fedoraproject.org alias
(‘official’ room) that can be added to the space, file an infra ticket.
Thanks for showing the way!
@nphilipp I guess official room in Fedora space will do. Let me know. Will raise a ticket.
Here’s the draft Wiki page for the SIG: SIGs/Music & Audio - Fedora Project Wiki
Let me know how you like it, or better yet, add what’s missing ! Once we have the communication channels established, I want to bring this out of draft and add it to the list of SIGs.
I did post some changes to the Scope section of the wiki. You all can see what you think.
Thanks! The first sentence got a bit long, so I made two out of it.
@nphilipp , I opened a ticket for a package review.
Finally, I gave up on JUCE. A lot of things in the fedora-review list.
I swtiched to jalv_select. A little GUI to select LV2 plugins.
https://bugzilla.redhat.com/show_bug.cgi?id=2251510
The name of the Matrix room needs to be lower case without special characters like &.
music-and-audio
music-audio
Which one do you prefer?