Does traditional sampling have a future?

[shamann]

Physical modelling is based on the concept of creating parameters to control physical attributes of an instrument to replicate its sound (like shape, materials used, sound excitation sources, etc), the most common variant relying on a filtered delay loop system, most famously exemplified in the Karplus-Strong algorithm.

Synful doesn't work that way. What steff3 is getting at with concatenative synthesis is closer, taking sample data and interpolating it with some form of abstract synthesis (in this case primarily additive). That has nothing to do with physical models.

[Roel de Witt]

Hi,

The only thing that matters is a reasonable size vs performance vs feel/control (vs price). The rest simply doesn't matter.

Terms like PM are very misused in marketing these days. Most products that claim to only do PM are in fact for a large part still using samples. In fact there is no VSTi out there that recreates acoustic instruments using PM/some form of synthesis without relying for a large part on samples.

---
Roel / 4Front Technologies

[Krakatau]

What steff3 is getting at with concatenative synthesis is closer, taking sample data and interpolating it with some form of abstract synthesis (in this case primarily additive). That has nothing to do with physical models.

this reminds me at a certain degree of one of the audio editors from IRCAM, called diphone studio...
...except that in the case of synful orchestra, the processes seems to be completely automated and strictly focused on its own library of instrument's phrases


Do i somehow get it ?

[shamann]

Do i somehow get it ?

Sure.

I was just trying to alleviate some of the confusion related around the concept of physical modelling, which I agree with Roel, is a much abused and overused term. All of which has little to do with sampling for the most part, except that hybrid forms of sampling/synthesis are pretty neat.

One thing I find interesting is the DVZ Orchestrator. It's sample library is quite small in comparison to things like Colossus, where most of its utility comes out of the extra systems. The system itself needs significant refinement for its methods to have any impact on common sampling applications (as it is, it is a highly niche novelty act), but I like the idea of these kinds of instrumental libraries focusing more on control methods instead of merely recording for every contingency.

[Bubbamusic]

You fellas have touched a nerve talking about "traditional" sampling - smaller and great quality samples/programs. Falling right in with this lovefest is THE traditional format: S1000 Akai. I still use it a lot. It seems to do a better job with stereo imaging and samples load and "audition" much more quickly than other formats. All that being said, it is quickly becomming extinct. Oldtimers like me still have large, albeit aging Akai collections from the hardware era. I still dust them off and use them. There is some great old stuff out there. I convert newer stuff to S1000 to make it more useful to me. It has to be a good collection, because converting to S1000 is one tedious bitch.I would hate to see the Akai format disappear.

Peace: bubba

[93143]

I wonder if that Ultimate Piano Collection would convert well to Emulator X...

Regarding traditional sampling vs. large multisamples, my Nearly Upright piano set will combine the, uh, best of both worlds! Programming that pushes the boundaries of Emulator X's capabilities, combined with a massive 4.4 GB sample set that expands to 6 GB when you import it! (Emulator X saves 24-bit samples as 32-bit.)

The big advantage is that no one's paying me either to get it done fast or to optimize it to fit on a ROM... The sfz version will be a challenge, though...

[chickeneps]

I wonder if that Ultimate Piano Collection would convert well to Emulator X...

I comment directly to this...

ULT PNO is tricky because it provides the S3000 programs without the samples in the same volume. It uses the samples out of the S1000 volumes so they don't have to write them twice, so on the Akai you have to load the samples and the programs in two steps.

I think the EmulatorX Convertor does do the S1000 ones correctly. If not, Chicken Systems has adapted Translator Free www.chickensys.com to convert these perfectly. We worked with this library comprehensively and we have it down.

[aMUSEd]

Physical modeling is probably going to be the future, but I imagine convolution as well as a limited degree of traditional sampling will be used. For example, as I understand it, TruePianos is mostly physical modeling based on a small amount of real samples. Synful Orchestra is PM based on a small amount of sampled phrases.

No, synful is in no means PM! it is concatenative synthesis and additive synthesis - no PM there!

best

OK, this is getting a little ridiculous. Synful is quite obviously not a synth that relies on a large sample library. Can we all agree on that? In fact, it is extremely small considering the range of sounds it creates, and the level of realism. The system integrates a form of sampling, drawing upon complete phrases (rather than individual samples) along with additive synthesis (as you said) and noise elements. I don't know if there's some academic definition of "physical modeling" but this system is CLEARLY not traditional subractive synthesis, nor is it traditional sampling. Unless you want to make another category for it all together, I think it makes more sense to just call it a PM system.

Not really. Physical modelling models how the sound is produced, not the sound itself. Of course if it models the way the instrument works well enough the model should be capable of recreating similar sounds but its the process that is the key thing. It is based on a model of the physical attributes of the instrument that creates the sound including how it is struck, blown, bowed etc. Some of the best trumpet and sax physical models for the Nord G2 for example only work if you input virtual "breath" into them (either using a wind controller or a midi control that can emulate the breath parameter such as modwheel). That is not the same as systems that recreate the sound - however accurate they are if they don't model the way the sound is produced they are not physical models. Resynthesis and other processes that break the sound down into components that can be used to reproduce and modify a sound (including Synful) are not physical modelling.

It's not true that most VST instruments based on Physical modelling use samples btw - Modelonia, Tassman, String Studio are just 3 examples - in hardware the Nord G2 and Yamaha VL70 also do not use samples, just synthesis (Tassman does use impulse responses as exciters for a few instruments but that is not the same - it is using the IR as a component of the physical model - not as a sample).

[Bubbamusic]

What was the skinny on the Yamaha VP-1 a few years back that was one of the first attempts at PM? They only built two units, they were mono and about $40K each. Is the VL series decended from the VP-1?

Peace: bubba

[Krakatau]

What was the skinny on the Yamaha VP-1 a few years back that was one of the first attempts at PM? They only built two units, they were mono and about $40K each. Is the VL series decended from the VP-1?

Peace: bubba

I don't think so, AFAIK, VP-1 was intended to emulate plucked strings and tuned percussion

Tassman does use impulse responses as exciters for a few instruments but that is not the same - it is using the IR as a component of the physical model - not as a sample

Sounds amazing that way , i'd thought sensible the placement of IRs at the end of the process chain ?
(...this beeing kept on topic by the fact that this technique can perfectly be applied to sample libray )