How to get around max start offset is 65536?

[Christoph Hart]

@d-healey yup, the 64bit size limit of the HISE event type says no.

Can you use the modulator? It's 0...1 value should map to the full sample start mod range of each sample. If you want to script it, use a voice start modulator in there that grabs the value you pass over using the event storage system.

[d.healey]

@Christoph-Hart said in How to get around max start offset is 65536?:

Can you use the modulator?

That might work, I will test and report back, thanks.

[d.healey]

@Christoph-Hart I don't think it will work in this situation. I'm porting over a Kontakt instrument -

In Kontakt this instrument uses the timemachine legato mode. So sample A plays, and then when you play sample B it starts playback from the current position of sample A.

If all the samples were the same length then I think the modulator would work, but in this case they are not so, for example, I don't know what 2000ms into sample A translates to as 0 - 1 of sample B.

Or perhaps there is a method I'm not seeing.

[Dan Korneff]

@d-healey Can you use getSampleLength() to calculate the 0-1 value?
Something like:

var totalSamples = currentSample.getSampleLength();

var offsetSamples = (offsetMs * sampleRate) / 1000;

var normalizedOffset = offsetSamples / totalSamples;

And then use the normalizedOffset value for your modulator?

[d.healey]

@Dan-Korneff how are you getting current sample?

[Simon]

@d-healey I added a function to my fork to get sample length based on note/velocity in the map, I've found it very useful.

[Christoph Hart]

@d-healey fair point. Could a method help that just multiplies the start offset for the sampler in question? This way you would loose a bit of precision but can yank the offset range into the dimension you need.

So if you need to start the sample at 100.000 offset which is beyond 65k, then you'll use

// Internally multiply the offsets by this amount
Sampler.setOffsetMultiplier(2);


// Later
local realOffset = 100000;
Message.setStartOffset(realOffset / 2); // 50k < 65k => all good

it won't be 100% sample accurate anymore, but below a multiplication factor of 8 this shouldn't have any real world consequences.

[d.healey]

@Christoph-Hart that might work, am I able to change it dynamically in on note on?

[Christoph Hart]

@d-healey yes but don‘t rely on the calls being always in order, if multiple notes are played in one buffer then it will use the latest value for the multiplier for all notes within that buffer.

[d.healey]

@Christoph-Hart Sounds promising, let's give it a try

[d.healey]

@Simon said in How to get around max start offset is 65536?:

I added a function to my fork to get sample length based on note/velocity in the map, I've found it very useful.

There's actually a built in way to do this.

Although I'm not sure if it's usable for realtime... but maybe I could "cache" this info during init and retrieve it in the on note on callback perhaps... something for me to play with

const var Sampler1 = Synth.getSampler("Sampler1");

const list = Sampler1.createSelectionWithFilter(function()
{
    return this.get(Sampler.Root) == 72 && this.get(Sampler.HiVel) > 20 && this.get(Sampler.RRGroup) == 1;
});

Console.print(list[0].get(Sampler.SampleEnd));

[Christoph Hart]

let's give it a try

You're chances of this not interrupting your workflow will rise by 100% if you try to implement this yourself and hit me with a sweet pull request. My local branch is heads deep into the rewrite for the new group management system and the implementation should be pretty easy:

All you need to do is to add the API method to the Sampler class and create a function in the ModulatorSampler::setStartOffsetMultiplier() class that stores the multiplier (ideally as an integer).

Then in the voice start code, you need to modify this line here:

https://github.com/christophhart/HISE/blob/961d7d903632982c4c315062271cf1318cee278b/hi_core/hi_sampler/sampler/ModulatorSampler.cpp#L1289

to include the multiplication with the factor.

[d.healey]

@Christoph-Hart Thank you

[d.healey]

Done - https://github.com/christophhart/HISE/pull/680