I know you hate it - They hate us too

[Innermost]

Strange. Being a twat is now a marketing strategy?

I exploit human reactance

[VOODOO U]

Does this work on Windows 10 1407?

Yes. Linux, Windows, MacOS

Ok great; I'm intrigued.

[Hans25]

Stop all this hypocritical AI datacenter save-the-world nonsense. All of you have purchased music hardware and phones produced in China, which is one of the world's three worst polluters.

When the market settles down, most of these datacenters will probably go out of business because they are betting heavily that everybody wants AI in the cloud. Furthermore, GPUs and LLMs will evolve, become more efficient, and eventually no longer need the massive power consumption they use now.

There have always been doomsday preachers, and they have all failed since mankind is still here.

I'm actually really impressed by what Suno can deliver. The problem is not Suno. The problem is that there are fewer and fewer real musicians with actual musical skills, and more and more advanced technology that takes the incentive to learn those skills away from people... including myself.

The songs that Suno created for me sound better than any new pop release I have heard in the last decade, at least to my taste. That says something about today's music. Suno cannot invent new genres, but it can deliver a solid, non-monotonous piece of music with a verse, chorus, bridge, etc., better than most of today's artists. That's my opinion.

[jamcat]

Interesting concept.

So rather than search Splice, it's leveraging some generative AI service like Suno?
Who actually owns the generative AI service that its using?
What are the terms of use? Specifically, who owns your generated data?

What are the limitations to its ability?
Is this just good for drums and other senseless sound loops that require little to no musical context, or can it generate a backing string section or choir that fits your song?

Also, basic question: What's the bitrate / resolution of the output?

It is also interesting to observe how people kind of choose the reception they get. The OP could have come in like a professional and pitched what this does and got a positive reception, but he chose to come in hot and basically say "fight me!" so everyone did.

Anyways, let's move the discussion now to actual functionality and limitations of the plugin and service.

Everything runs locally on your CPU. Your data are yours.
Under Stability AI license. Since you win less than one million a year, you don't have to pay anything at stability ai.
Bitrate depends of your daw. Personnaly 48khz. Stable audio model generate at 44100 but the VST upsample or downsample in function of your settings.
Personnaly I love generating any kind of sounds. From drum, to guitares, synths, voices, fx... In fact everything depends of your creativity and your ability to hack the model.
Generated datas are yours.

OK, so let's see if I have this straight:

If you sign up for the beta, you get a lifetime license of Obsidian Neural Local Edition, which does not expire at the end of the beta period?

The Local Edition leverages Stable Audio 3.0, which is (currently) a free service for losers under one million a year?

Obsidian Neural provides essentially an interface that communicates between your DAW and Stable Audio 3.0, which can take text or audio(?) as token data for generation?

Stable Audio 3.0 runs on your local machine, which you can get by itself from stability.ai and get the same processing as Obsidian Neural Local Edition, just without the DAW plugin middle man?

Assuming this is all more or less correct, how large is the Stable Audio 3.0 model that you download to your local machine?

What is the audio resolution (bit depth)? 16-bit? 24-bit? 32-bit float?

[Innermost]

Interesting concept.

So rather than search Splice, it's leveraging some generative AI service like Suno?
Who actually owns the generative AI service that its using?
What are the terms of use? Specifically, who owns your generated data?

What are the limitations to its ability?
Is this just good for drums and other senseless sound loops that require little to no musical context, or can it generate a backing string section or choir that fits your song?

Also, basic question: What's the bitrate / resolution of the output?

It is also interesting to observe how people kind of choose the reception they get. The OP could have come in like a professional and pitched what this does and got a positive reception, but he chose to come in hot and basically say "fight me!" so everyone did.

Anyways, let's move the discussion now to actual functionality and limitations of the plugin and service.

Everything runs locally on your CPU. Your data are yours.
Under Stability AI license. Since you win less than one million a year, you don't have to pay anything at stability ai.
Bitrate depends of your daw. Personnaly 48khz. Stable audio model generate at 44100 but the VST upsample or downsample in function of your settings.
Personnaly I love generating any kind of sounds. From drum, to guitares, synths, voices, fx... In fact everything depends of your creativity and your ability to hack the model.
Generated datas are yours.

OK, so let's see if I have this straight:

If you sign up for the beta, you get a lifetime license of Obsidian Neural Local Edition, which does not expire at the end of the beta period?

The Local Edition leverages Stable Audio 3.0, which is (currently) a free service for losers under one million a year?

Obsidian Neural provides essentially an interface that communicated between your DAW and Stable Audio 3.0, which can take text or audio(?) as token data for generation?

Stable Audio 3.0 runs on your local machine, which you can get by itself from stability.ai and get the same processing as Obsidian Neural Local Edition, just without the DAW plugin middle man?

Assuming this is all more or less correct, how large is the Stable Audio 3.0 model that you download to your local machine?

What is the audio resolution (bit depth)? 16-bit? 24-bit? 32-bit float?

- lifetime license yes.
- one million yes. Check the stability ai license on their website. When you install my VST you have to sign the license.
- yes that's right. The vst is the interface which communicate with the audio model. But the VST makes the model run via onnxruntime
- yes if you wanna make run the model without my VST possible too. Model is open weight on huggingface. Search for innermost47 on huggingface and stable audio 3 onnxruntime model, I have the model on a repo. And Stability AI host their model too on huggingface.
- stable Audio model is something like 7gb
- don't remember for the bit depth

You didn't ask but the model is on the ONNX format in order that it runs on your CPU

[Funky40]

You hate AI?
Proove it.

No, i have no problem with AI.

The "problem" is your webpage does not give out the info easily. And this would be then your "problem" not the mine.

Your shown attitude is something else. Some might not care, some might.....


edited out

[Innermost]

You hate AI?
Proove it.

No, i have no problem with AI.

The "problem" is your webpage does not give out the info easily. And this would be then your "problem" not the mine.

Your shown attitude is something else. Some might not care, some might.....


edited out

Yep. You're right. Some will not love me. Some will not care. And some will like it. Every choices has its consequences. I made mine. And you, what is yours?

For the web page problem thanks for the feedback. Will think about how to make it clearer. Or not. Lot of jobs on the VST itself for a solo guys who don't use AI agent to code.

[Innermost]

What is funny is that every forum discussion seems to drift at a moment.
I prefered to drift first.
We are so predictable.
Perhaps that's why AI works... It sees our patterns. And reproduces it so well that we recognize ourselves in it.
And perhaps that's why we are such afraid about AI.
Because we are afraid about us.

[dionenoid]

@mods : Please move this to the correct subforum : viewforum.php?f=316

[Innermost]

Please move this to the correct subforum : viewforum.php?f=316

Why? I am speaking of a VST instrument.

[Aloysius]

The people of Holland think your thread belongs in ''Machine Learning and AI for Music Creation''.
No instructions were given on how to move said thread to that location.

[Innermost]

The people of Holland think your thread belongs in ''Machine Learning and AI for Music Creation''.
No instructions were given on how to move said thread to that location.

What other people from other countries think? And you?
Furthemore, based on what they think that?

[Aloysius]

The link leads to ''Machine Learning and AI for Music Creation''. Since OBSIDIAN-Neural uses AI for music creation, the Thread would find a natural home there.

[Innermost]

The link leads to ''Machine Learning and AI for Music Creation''. Since OBSIDIAN-Neural uses AI for music creation, the Thread would find a natural home there.

Ok, so as it uses AI with all the features that I dropped into the conversation you still don't recognize it as an instrument.
Well, then, what is an instrument for you?
Perhaps that would be better that we find a good definition for this word then, on the contrary one of us could feel upset.
And while we are debating some guys on the forum click on the link and get my beta for free. A lot since I dropped this post in fact.
Thanks guys.

[dionenoid]

I think i speak for most when i say that i get sick and tired of all these AI discussions all over KVR. So i asked the @mods to move this topic to the correct subforum.

Seems to me a very fair request, seeing how the dev started this topic as an AI discussion, and especially seeing most replies being the usual Hyde park type replies about AI.