What happened to the SmootHound Siva?

Fwiw, here's Leon's video on the Ampero. You can easily see how flawless the editing basically works. Comparing this to what you can see in the Siva video is, uhm, well... just impossible.
 
It would be cool to invite Smooth Hound's Chris Fryer over here and maybe do a AMA of sorts 🙏 He used to lurk on TGP if i recall correcty.
 
Why exactly? And yes, I'm genuinely curious. What exactly peaked your interest? There's barely any sound clips, let alone decent ones, the website is looking like a shady business, the 8ms of latency you get from his wireless system aren't exactly a great "buy Smoothound!" ad, either. So what else is it?
I like the concept. I like the technology - SPICE style models at electronic component level. I like that the supposed latency is less than 1ms. And tbh I really like how it sounds in the soundcloud clips. Remains to be seen how it feels though but considering that the device is supposed to mathematically model the circuit I suppose it's promising.

And by the way this is an absolute departure from what people call "component based modelling".
 
I like the concept. I like the technology - SPICE style models at electronic component level.

Yup, it's very impressive. As far as i know Fractal (and maybe HeadRush with Revalver?) are the only other company out there creating models this way.
 
Laxu, Sascha and Mbenigni when they land in a UI thread…

Ice Skating Marriage GIF


LFG
 
I like the technology - SPICE style models at electronic component level.

Isn't that more or less what pretty much all modeler makers are using?
I like that the supposed latency is less than 1ms.

That'd be impressive but where does it say so?

And by the way this is an absolute departure from what people call "component based modelling".

How so?

Admittedly, the SC demos sound pretty good. But it's still not that other modelers wouldn't sound as good.
 
Isn't that more or less what pretty much all modeler makers are using?

tl;dr, no. Most "component" modelling products out there actually model different signal path sections (low-pass filter, gain stage, etc.) which then get connected together. This can yield very accurate results with relatively low DSP overhead, but requires a lot of manual work in the form of circuit analysis and live measurements.

SPICE involves modelling the actual underlying circuit, end to end - and doing this for audio in real time is no easy feat. It however means that you can more or less dump a schematic in and get a functioning model out, with little tweaking.
 
Isn't that more or less what pretty much all modeler makers are using?
Not really.
That'd be impressive but where does it say so?
I think it was the first yt video he posted.
Just like that. I posted some information in the Tonex thread but people would rather use made up terms to sound smart.
Admittedly, the SC demos sound pretty good. But it's still not that other modelers wouldn't sound as good.
Not an argument in favor or against.
 
tl;dr, no. Most "component" modelling products out there actually model different signal path sections (low-pass filter, gain stage, etc.) which then get connected together. This can yield very accurate results with relatively low DSP overhead, but requires a lot of manual work in the form of circuit analysis and live measurements.

SPICE involves modelling the actual underlying circuit, end to end - and doing this for audio in real time is no easy feat. It however means that you can more or less dump a schematic in and get a functioning model out, with little tweaking.
I’m having a very hard time reconciling this with the alleged 1ms latency. Is the SIVA running a massively overpowered processor in comparison with e.g. FAS, etc?
 
All Boss devices are 1ms or less, it's not impossible.
Neither here nor there. The specific point I’m addressing is that granular, SPICE-level, per-physical-component modeling described above is more computationally intensive than the algorithms used by competitors. You wouldn't expect to take this on and arrive at significantly lower latency. I'm not saying it isn't possible; I'm saying that another variable has to move in one's favor (hardware, serious optimization, etc.), and I'm wondering what did. (Or whether something here hasn't been fudged/ misrepresented.)

 
Back
Top