How Google A.I helped me tune my new audio-system!

[Hindsider85]

I bought the Fosi V3 Mono in a complete set from Ebay and got them today.
Im using the Nubert Nuvero 70 speakers, from the venerable German audio specialist "Nubert".
I have the Hifiman Serenade with the Gold Upgrade and use it as a Pre-Amp and DAC, to feed the Fosi V3 Monos over the internal Streamer.
Im using Tidal Streaming and Audirvana as a pre-processor, for the upsampling to DSD 512 via r8Brain.
I gave the Google A.I lot of parameters and the sound I was trying to achieve. Since the Hifiman Serenade is a pure NOS R2R DAC, I knew it would be benefitial to use Audirvanas r8Brain upsampler, to achieve the best fidelity possible. My PC is a High-End system and the processing power, albeit hefty, is no issue at all.
The Google A.I gave me the settings and they differed starkly, in some areas, from my own.
And what can I say...The A.I was right from the get go and I didnt want to believe it at first.
I was always using 85% bandwidth with the Typ B - 4th order, 160 dampening and minimum phase.
It told me that I should use 90% bandwidth, Typ A - 4th order and the rest could stay the same.
And it was absolutely right.
I got a very mellow sound, without sacrificing stagewidth and fidelity of the higher end frequencies.
My settings where rolling off to early (18khz range) and the Typ B -4th order was causing background details to vanish too quickly. Like a harpe vanishing too abruptly and not slowly rolling out. And the staging was compressed, which I only realized once I used the settings the A.I provided. Suddenly it opened up, dramatically!

I know hating A.I is the new "vogue" thing for many people, but if you use it correctly it can solve many issues in your day to day life.
Surely your shouldnt take anything it says for gospel, but, given enough parameters, it can quickly and correctly give you perfectly viable information in a mere second.
Truly impressive times we are living in.

 

[sam_adams]

The Google A.I gave me the settings . . .

Did you ask it; "Are you sure?"

 

[Hindsider85]

Did you ask it; "Are you sure?"

Why should I? I simply tried them out and those did exactly what it said it would. I listened to it extensively and made quick A-B comparisons, which is quite easy with Audirvana.

 

[sam_adams]

 

[Chrispy]

How do you limit bandwidth?

 

[DVDdoug]

You can't trust AI for audio. AI is just reading & repeating the "audiophile nonsense" that dominates the Internet and publications. This is one of the few scientific-rational resources.

Rolling off at 18kHz MIGHT be audible but it's a silly thing to do... Otherwise, I don't think you made any changes that would affect the sound. You didn't say anything about EQ. Nothing you are doing can affect "soundstage" or "width" unless it's a side-effect of roll-off or EQ. But in that case you'd perceive the frequency response changes more than any side-effect.

Up-sampling doesn't change the sound when done properly. Down-sampling shouldn't change the sound either as long as you stay at "CD quality" or better.

and made quick A-B comparisons,

Doesn't count!

Controlled Audio Blind Listening Tests (video)
What is a blind ABX test?

Audiophoolery is also worth reading. He talks about the FEW REAL characteristics of audio. Audiophiles have a dictionary full of undefined terms like "detail", "clarity", "musicality", "resolution"* etc., etc., etc. Audiophoolery can help you to ignore all of that and stick to what's real and defined.



*Resolution has a meaning in digital audio (and video) but it doesn't really describe the sound. If you make an 8-bit audio file you'll hear quantization noise in the background. THAT's what low resolution sounds like.

 

[Jiraya369]

You can't trust AI for audio. AI is just reading & repeating the "audiophile nonsense" that dominates the Internet and publications. This is one of the few scientific-rational resources.

Rolling off at 18kHz MIGHT be audible but it's a silly thing to do... Otherwise, I don't think you made any changes that would affect the sound. You didn't say anything about EQ. Nothing you are doing can affect "soundstage" or "width" unless it's a side-effect of roll-off or EQ. But in that case you'd perceive the frequency response changes more than any side-effect.

Up-sampling doesn't change the sound when done properly. Down-sampling shouldn't change the sound either as long as you stay at "CD quality" or better.


Doesn't count!

Controlled Audio Blind Listening Tests (video)
What is a blind ABX test?

Audiophoolery is also worth reading. He talks about the FEW REAL characteristics of audio. Audiophiles have a dictionary full of undefined terms like "detail", "clarity", "musicality", "resolution"* etc., etc., etc. Audiophoolery can help you to ignore all of that and stick to what's real and defined.



*Resolution has a meaning in digital audio (and video) but it doesn't really describe the sound. If you make an 8-bit audio file you'll hear quantization noise in the background. THAT's what low resolution sounds like.

you cant even trust people tbh lol

 

[Chrispy]

you cant even trust people tbh lol

But especially not the ones who have disseminated so much nonsense about audio for the AI to glean....

 

[Jiraya369]

But especially not the ones who have disseminated so much nonsense about audio for the AI to glean....

as if people here dont spread nonsense either lol it's not the AI's fault people in the audio world all spout bullshit

thank God for Dr Toole.

 

[Chrispy]

as if people here dont spread nonsense either lol it's not the AI's fault people in the audio world all spout bullshit

thank God for Dr Toole.

People here spread little nonsense generally....some not so competent audiophiles sneak in sometimes, tho. Just no reason to continue much of the audio bullshit via AI

 

[amirm]

I ask Gemini to opine on your post. Here is its final conclusion:

"

Verdict​

• Is it true that the settings changed the output? Yes. Changing bandwidth and filter types creates a different measurable analog output.
• Is it true the AI "knew" better? Likely not. The AI simply suggested a less restrictive filter set, which generally sounds better to people who prefer a "natural" or "NOS" sound profile.
• Is the "dramatic" improvement real? Without a blind ABX test, most experts at Audio Science Review would attribute the "dramatic" part to confirmation bias."

 

[DVDdoug]

Excellent!!! Gemini is referencing and respecting ASR!!!

 

[tineared]

Why should I? I simply tried them out and those did exactly what it said it would. I listened to it extensively and made quick A-B comparisons, which is quite easy with Audirvana.

Sure. But the adjustments that Gemini convinced you to do and subsequently also convinced you that they are the droids you've been looking for, also initiated an embedded spying subroutine which will track your every keystroke and reveal all that is known about you. No secrets.

Happy listening!

 

[Jiraya369]

People here spread little nonsense generally....some not so competent audiophiles sneak in sometimes, tho. Just no reason to continue much of the audio bullshit via AI

little? lol

 

[DavidMcRoy]

I've been consulting Google Gemini AI in designing and constructing a new large quasi open baffle speaker. I've found it very useful, with caveats.

I've determined that it's most effective to ask it the same questions over a period of days or weeks in search of some kind of consensus among its responses, since they change.

It is also subject to factual errors, which it will acknowledge when called out, and alter its recommendations when newly enlightened.

AI is what it is, GIGO...garbage in, garbage out...but taken in context is a useful tool.