Valo - public beta demo, second attempt

[mystran]

Ok, so announcing second attempt at a public beta demo, for my PC-VSTi synth Valo. (First beta was in this thread.)

Demo-limitations as before: 1 instance, 4 voices max.

Still only SSE build, I'll need to hunt a few denormals down before I can release a non-SSE, and I'll do that after stuff seems a bit more stable.

Will post a new screenshot in a day or two.. until then you have to grab the new plugin.

New stuff:
- pretty much complete modulation rework, more flexibility, more orthogonal
- 2 more LFOs (+2 extra routing slots) and the ability to combine LFOs, with more targets
- single-cycle, tempo-sync and song-position-sync LFO modes
- mostly relayout interface, stuff went a bit smaller to keep GUI to sane size
- no new graphics yet (only very minor tweaks to the old ones)
- delay effect (normal and ping-pong for now, separate initial/feedback times, stereo-offset and/or pan)
- basic brickwall limiter, 0.5ms lookahead (hopefully not too much to prevent live play), can be mixed with (latency compensated) dry signal for lesser ratios
- some odd stuff fixed
- host automations done for the old version will break! (sorry, will try to avoid that from now)

Old stuff:
- people seemed to think it sounds good, hopefully that hasn't changed
- old presets should work; the converter isn't 100% accurate (too many changes) but it should get things very close
- same old download location

This is named "pre1-pb2" 'cos this is fresh enough that I didn't extensively test everything after the last build. Everything has been tested during development though, so hopefully there aren't any really bad bugs.

In any case, I'm calling this "unstable" in the sense that there might be little (unsupported) changes and tweaks here and there. Big changes that really break patches will still get converted as necessary. But this might not be a safe version to rely critically on. Some info from the display is also still missing, but since that's independent of everything else, I'm letting this live anyway.

Apologizes for anyone that thinks the new interface is more intimidating than the old one. Once it seems the basic interface is going to stay stable (hopefully now) I'll see if I can get somebody do some fancy graphics.

Modulations work like Output=LFO*(Amount+Source*Mod) with the only tricky thing being that disabled "Source" will act like a constant 1 (making "Mod" just an extra control for Amount, this is mainly useful for the extra routing slots, where there is only one "Amount" and it works like "Mod", but I made it work with LFOs as well).

Feature requests and complains are still welcome, but I probably don't do any drastic changes again (but you can always ask anyway.. pfft).

On TODO list:
- full version shipping (as early adopter version anyway)
- minor sound enhancements (especially the effects I'd still like to tweak a bit)
- interface enhancements (hopefully mostly "minor" stuff, but anyway)
- more presets (current set doesn't show that much of what can be done)
- sequencer (at least gate, but probably notes too), maybe arp (scheduled 1.0)
- real artistic graphics (probably not before 1.0)
- full mod-matrix (scheduled 1.5 or 2.0 if we ever get there, but I'd like to do it eventually)
- more osc-waves, more filters, more LFO waves, maybe even more targets


....

feedback?

[mystran]

Do people think we've improved, or should we go back to the old version?

[John-E34]

Stick with the new version, not much point going backwards.... personally, I think the target audience has had SE/SM stuff 'up to the eyeballs'.

You now find warnings on many developers sites stating the 'problems' with Synthedit developed vst's. Maybe Synthmaker is ok, but wow, so much stuff released....

Who really has the time, when at the end of the day, when dev's get really good at what they do, they abandon their customers and go work for the big guys......

Not suggesting you are in the same boat, just airing my thoughts

[mystran]

You now find warnings on many developers sites stating the 'problems' with Synthedit developed vst's. Maybe Synthmaker is ok, but wow, so much stuff released....

Hehe, well pure old fashioned C++ all the way here.

[John-E34]

I stand corrected... sorry for that

[mystran]

Forgot to mention, I didn't include any of streifentier's patches in the factory bank, even if I had the permission (and there would be tons of room currently), 'cos thought I'd give him a chance to check that they convert properly first and if he wants to tweak or add something now that more options are possible.

[stanlea]

Hey mystran the new version is just fine, just a problem with the gui black letters on grey background are hard to read.

[mystran]

Hey mystran the new version is just fine, just a problem with the gui black letters on grey background are hard to read.

I take this is with the default colorscheme (now "Light" is default). There's enough contrast on this TFT, but maybe you're using a traditional CRT? Guess I have to increase the contrast some amount. Thanks for mentioning this.

In any case, there's more colorschemes (including the original) accessible by right-clicking somewhere on the background (the Valo-logo if safe place for example). Maybe one of them is better for your eyes?

[mystran]

There seems to be somewhere a nasty (probably in the GUI code) that causes it to sometimes (but not too often) crash. If you see this happen, please describe what you were doing when it happened. Even vague information is appreciated as I want to find this one quick.

[edit] Ooops.. found one... uploading fixed version in a minute... hopefully not more, but still report any you find... [/edit]

[mystran]

Hey mystran the new version is just fine, just a problem with the gui black letters on grey background are hard to read.

Would this be easier to read, or is the difference insignificant:

[mystran]

Not sure if I caught the occasional-crash bug, but I hope so.. and I uploaded a version with some likely causes fixed; identifies as pre2pb2, otherwise identical.

[stanlea]

Would this be easier to read, or is the difference insignificant:

Better, thanks

[bwwd]

I like the sound , it sounds nice , multisaw is a supersaw ?Stringy patch is nice , but as first and most important improvement would be CPU eating , i have 50% cpu resources taken when i play 3 notes on stringy patch it takes meoff a little bit.And sometimes sound is a little bit "muddy" i dont know how to desctibe it.But im impressed it sounds really good but this is little cpu hog for me.
Man i LOVE finetune of oscillators:) please make it good for cpu and add some optimisations:)
Filter amount is alsoo needed , i cant get good plucky sounds from it without it.
Also i would like to have yellow gui with red elements but not too bright.

[mystran]

I like the sound , it sounds nice , multisaw is a supersaw?

I'm avoiding the wording "supersaw" because I'm not sure if certain companies consider that their trademark wave by now, or if people assume things if you say it's a supersaw. In any case, multi-saw is 8 independent, detuned saw-waves (with independent instabilities and all), mixed together and scaled down by average estimated total gain. The detuning used is a static guess for a set of detuning ratios to minimize "beating" (ie. that the phases get "random") after the 8 saws start from the same phase (as with the reset version).

So I guess it's somewhat similar to what is known as a "supersaw"

Stringy patch is nice , but as first and most important improvement would be CPU eating , i have 50% cpu resources taken when i play 3 notes on stringy patch it takes meoff a little bit.

Would you care to mention the CPU type (and core frequency, and samplerate)? It does eat about 25% (of 1 core, and with 3 notes "Stringy", 44.1kHz sampling) of this AMD TurionX2 clocked at 1.80Ghz, so I admit it's not the lightest thing around. This was designed for a certain sound-character and quality, and I knew from the beginning that it's not going to be the cheapest possible on the CPU, but I'm looking to improve the CPU efficiency as soon as other things permit.

The most expensive single component is the filter (this is profiled to be a really significant part of a voice-rendering time), since the model is highly non-linear and therefore rather expensive to calculate without aliasing. I'm looking to maybe do a cheaper, more "modern" (and cleaner) sounding filter (as an alternative) at some point. There might be optimizations possibilities with the current filter as well, but it's pretty tight all ready; it just does a lot of work to get it sound like I want it, and any further optimizations will require pretty low-level analysis of bottle-necks or some really clever transformations. There are some possibilities though, and I'm looking to take advantage of them eventually.

Then there are some optimization possibilities by skipping unnecessary modulation calculations and other stuff like that (I'll do it as soon as I find the time, but it's slightly non-trivial so I need to find a few hours to figure it out properly at some point). Some of the effect code is also new, and therefore didn't get much work on CPU efficiency yet. Finally a lot of parameter-info is still re-parsed for each block, which will give an unnecessary CPU hit when using short processing buffers for low latency (and this also is thing I want to fix as soon as I find the time to go through it all).

Finally, I'm looking to greatly reduce the processing time for the chorus, which is currently needlessly CPU costly at least on my CPU, considering what it does anyway.

And sometimes sound is a little bit "muddy" i dont know how to desctibe it.But im impressed it sounds really good but this is little cpu hog for me.

Well, it can get a bit muddy when driving the filter hard (mixing 2 oscs 50%+50% vs. 100%+100% does make a difference) especially with highish-resonance because of the non-linearities. This is by design. If incoming waves are further detuned from each other, this will cause inter-modulation distortions (depending on how hard it's driven) which can be reduced by lowering mixer gains if desired. Again, this is by design. Maybe in the future we'll have a "modern-softsynth" style clean signal path available as well, but then it won't sound any special anymore.

The oscillators and filter are also somewhat "unstable" in a way that inevitably adds some noise to the output, and can make it a bit muddy with certain settings. On the other hand, this is also what gives them the character they have. It is on TODO list to add cleaner processing as option for sounds where this is not desired (the core oscillator model is basically "totally clean" if tuned that way). However this slightly noisy, breathing character was the original design target (and I actually go to the pain of "shaping" the square/triangle from the unstable-saw, and then shaping the "sine" out of the triangle to preserve this character).

The other thing where it can get a bit muddy (in a not-so-nice way), is the chorus, which admittedly could use more transparent interpolation. Cleaning the chorus (and other similar improvements to the other effects) will hopefully happen as well as time permits.

Man i LOVE finetune of oscillators:) please make it good for cpu and add some optimisations:)

You love finetune, or the fact that they aren't quite stable?

Like said, I will do what I can do improve the CPU efficiency, but this might not happen overnight, since the trivial (and some non-trivial) optimizations are mostly there already. But I'll do my best to get it down further, by hopefully getting more out of SSE and stuff like that in the future.

Now if I could manage to find the time to do everything I want to do.

[mystran]

Would this be easier to read, or is the difference insignificant:

Better, thanks

Will put that shading in the next version then.