Time Stretching

A Former User

//processFrame is the audiobuffer 1 sample at a time:

template <typename T> void processFrame(T& data)
		{
			for (auto& sample: data)
			{
				sample *= .5; // Volume @ 50% 
				// do other stuff
			}								
		}


//you can also do it this way (the method directly above processFrame):

		template <typename T> void process(T& data)
		{
			static constexpr int NumChannels = getFixChannelAmount();
			// Cast the dynamic channel data to a fixed channel amount
			auto& fixData = data.template as<ProcessData<NumChannels>>();
			int numSamples = data.getNumSamples();	//data also has some of its own methods you can call, like getNumSamples()

			for (auto ch : data)
			{
				dyn<float> channel = data.toChannelData(ch);

				for (auto& sample : channel) //For each sample, same as processFrame
				{
					sample *= .5; // 50% volume					
				}
			}

			//now we don't really need this stuff, since we did it manually:

			/*

			// Create a FrameProcessor object 
			auto fd = fixData.toFrameData();			

			while (fd.next())
			{
				// Forward to frame processing
				processFrame(fd.toSpan());
			}
			*/
		}

d.healey

@iamlamprey Thanks! How did you figure it out? Any idea why my node won't show up in HISE?

clevername27

@Christoph-Hart If you're really interested, I could arrange something with zPlane.

d.healey

@clevername27 Elastique would be a good addition! We'd still need Rubberband as a baseline because the other algorithms don't have compatible licenses.

clevername27

@d-healey Hit me up if you want to discuss further. That goes for any other commercial music software company, as well. It appears I'll be using HISE here, so I'm now invested.

d.healey

@clevername27 said in Time Stretching:

I'll be using HISE here

Good choice!

A Former User

@d-healey said in Time Stretching:

@iamlamprey How did you figure it out?

Christoph told me lol

A Former User

@d-healey said in Time Stretching:

@iamlamprey Any idea why my node won't show up in HISE?

If it compiles okay, and doesn't show up in HISE it means there's an error in the code somewhere that the compiler didn't pick up, or you haven't linked the library(s) properly

d.healey

@iamlamprey Now I don't feel so bad :p I'm trying to copy what the audio programmer does in the video but he's working directly with JUCE audio buffers. Do we have any access to them within the process function?

A Former User

@d-healey check out this thread:

Block-Based Processing in Third Party Node

Sorry for the dumb question, had a bit of brain fog the last few days template void process(T& data) { static constexpr int NumChannels = getFixChannel...

Forum (forum.hise.audio)

d.healey

@iamlamprey Thanks, that got me a little further :)

@Christoph-Hart Any idea why my node won't show up in HISE?

d.healey

@Christoph-Hart I have it working in Windows, but it still won't show up on my Debian machine. The DLL compiles successfully but it won't show up in the node list and I see this in the hardcoded fx

d.healey

Aha I solved it. ~~I was linking against my distro's rubberband library. I had to compile it myself and link against that instead.~~ actually looks like I can use my distro's library, if I move it into the src folder and point the auto generated jucer project to it.

Christoph Hart

@d-healey There's also a Rubberband build configuration where you don't need to link against a prebuilt library and it builds it from the source files directly. I used this method and just threw the entire Rubberband library into ThirdPartyCode/src/Rubberband...

d.healey

@Christoph-Hart Is that a configuration on the HISE side or within the custom node?

Christoph Hart

@d-healey In the Rubberband library

Single-file build. If you want to include Rubber Band in a C++ project and would prefer not to build it as a separate library, there is a single .cpp file at single/RubberBandSingle.cpp which can be added to your project as-is. It produces a single compilation-unit build using the built-in FFT and resampler implementations with no further library dependencies. See the comments at the top of that file for more information.

Although I think I managed to switch to IPP despite what this paragraph said...

rubberband/COMPILING.md at default · breakfastquay/rubberband

Official mirror of Rubber Band Library, an audio time-stretching and pitch-shifting library. - rubberband/COMPILING.md at default · breakfastquay/rubberband

GitHub (github.com)

d.healey

@Christoph-Hart Oh interesting, thanks

d.healey

@Christoph-Hart Yep that worked, much simpler :)

d.healey

Okay here's what I have so far. Mainly taken from the audio programmer's video, with a little help from some forum members and the RubberBand docs.

It certainly affects the audio but doesn't seem to be controlling the time stretch properly. Just makes everything sound kind of choppy. I tried it with adjusting the pitch instead of the time and that seemed to work much better (still choppy though).

I'm sure I'm missing something important. In the audio programmer's video he's using read and write buffers, but I don't know how to access them in a custom node (do we have access to them? are they necessary?). I'm just using the data pointers.

I also wonder if I need to add some start padding and delay as described here - https://breakfastquay.com/rubberband/code-doc/classRubberBand_1_1RubberBandStretcher.html#a078827068efc609efb4c71bd094d80e9

Any help appreciated.

rubber_band_node.zip

blezzbeats

@d-healey Excited to see where this goes. I've done some experiments abusing the timer, here's an poor man's old school timestretch:

HiseSnippet 2440.3oc2ZssaabbFdWKttgzwEMoBEs2MUn0cYprDojkRRUULsN4HDqCQT0tWXffQ6NjbhWtCwtC0gDHffhh9Zzq6M4YnnW0K5KQAJ5UEn4Mn8+elYOQtThV0IF09Bat+Gl4a9Ouy5CiDdr3XQjkc0iuX.yx9scZeQnr2l8n7Pqc2xx9tNRdeVrLhI85YswECnwwLeKa6YdLJhc0JVp+7MObCZ.MzikQxx5oBtG6I79bYF0Ca8I7ffcn9rigENi9CZsqmHbSQfXH.mYbZXMf58BZW19TTra4XYe6s84RQTaIUxhsrqrgv+h18DmEpk+o7X9IAL7glVsgERSdGQfOhXjp0l83A9GlbrisrrcNLyHLi1HLqydbedJ8LiwOPwfjoQd6g8sJBuYJ.ul4gWibvqDHYmCRNZH8NNs8h3CjYbP7bGmcCkrnNTvrmGJ6Ore6.tOK5PvDBPww9VfEVuBV1+EmMEfZgxE5SeAamH3gzkwc0FMlm.+U80pUqyvPOIWDRDg6KjrCBcqW6KqUs1k0HixpSmR4gaTjHHfEUJaLBH5pTzMbX+SXQySNkFLjkJH36KZqclrsNenfm9bmSPQ3tgb4ACXgSJ.wxXrfe8a1cKpjhNHCMPtArHIGgf8VrSgncs6ppyVr3WHECf38LeYEMSHJR3OLfJKFZg4SFFfMnf+DcOgwb4E4y2FMdqxzEuc0oCSKbeGmC4PAgxw6sJAufU6aa7Zxduqy1c5v7jYfshyN+1qLUsxzE9L8P4Gl.kGMzmKdh.CSTPYVMEBRhbX.8BVTyIUUkr7e5eLsUUsl5ppP8cOvswxGJo7kGGAEK3gcyy.g41gTvV3mGAGIDRLsOkVqOsUaZ+AALvhFIgkufzrSYQwEWgc3AlB6+Bms+UOGLceN3xhe9wa7wOeCFUR1fJkArmCtGobgynmZ0GsKN1nx8omi+rxezxJ..nZOS3EnPreBe6YtQEJx6oO.P+ob1YYzJsnckWC8QZN88QpbU8QtqywSnIuomws9muaZSiXl7YbeYO2UVF6Rji7Gy3c6IcWQ07XwEeLSRnpnczoPFnh1qAgiwRnddDIM2vjGPVmnF9XgtLohmNjJEptyMdtyb5sBKhShUhSDc5.XIlzQDQDfuKfNX.DVaXGWKhk7aUfySoAvF2XsRnuDvXDI+kiPXwkzG0Hv0Rh4eAi31GRD4wL3b5GWWspcQtsQlqSVogRg1ZaMg1WLLTRbg5CdfUDF2QqhwW7n9RPmlKzvbJgtzPmPxIWPNqGGTWJH7PuHFMlU73SfEj.KQDuaWvniKIOQ60y.z6ksOY1QHcEfztaoMUpm10uAhi0xSoolhhTDTPHDJH.jV0XJy8nIVP1KAjQPjMwsixBfT0QIcfZBDZnOo63B2IRzGnAEF3P3Ci5WOm+RIw5jsC6xCYXziNv4HfracCB6SCGpN6xngLEjvhXJTZBIiLE0.YRhoo99eRn3D24hPFyMOAlMBCtSDEC5cmSxNWB7lKcAmaLQ5CEbQQdB.PZz37giz.ze.x.N6w2Bn3Gv5AkvfdNv3Cd.dL4gAvxmeJpDYSllxSzefHDNY4GnJmuSQbsZWBKVhI.rlalnUhcnNt2l0bSZPvIPiC2w1MDR4hzyhvIu.roj6kBTi82DJVl42rFn0YYiKvHcQOPhbixWa9ZrPiliwQY+VRayGQGeen3i.FiAvXlionPF+59hn9z.vA5eX5wbLT3y5PGFfUMFlyQWhayn0U503cb0gzquNFQWuVUfXUPkXQ.agAQP1tqV50.5EJmX7xUurVUV.T3H2R0gBDLqkGMvCGtiYrpXpDpy0VGo53UQGM0LdGQzdXcx155jtIKZcrf5ji+RcvkGAVzvoaMrYxwHM.ymAi66isD.qsJx2ssnOS1CIcF75PcqOlSYbigwInz+ifxbFqVAK8dTna1.wYtP.172W8D8jXkN2OIsq9hMAt0IuW9hYKl6ghdJ098qml86+4sqjMP0IB3eIZY2SWREZyR8+7gwRz7AMjFfCSZxo2KopatT5MFJkhPL0C4gYeqB4zMwbAs3Eyo0zlq7LEMyqKQ4mZdpp5.UMsUfJZGSOROviHftWAxexgjliwDhHK.v5iNqP4EBSypvhekUNLU.v9r7GZJHVPqhlvrcbtxEb71SkIjpH5RqLAt5BoMl.TFo1mZUJweVPyqzslelpjBZE2WudzvtLeUYqWm0rx7VSHForCsNT4X83aD7J3hFOPAGh1Lgm5lTx.MNRnT8pfQGbB9lUnUHunHRT+HEG44tINTJKpLeTIhkTKj3N5X0vb0lxT4lhTOiO91.3vBOiK6oT3.0bqtMmWURFJHrz6O+HCaqakM5n6iMg9z63VarVliugY0dG678QjFj6cuQA.TbdkFMdC+fu5JqjdDK6ko9+6C9kS3lO2CdOTXxtE3cCEQrswSnqZ5KPKyHzIh.XA0be08WpdKjpuLyYMEUrJY7sbSBVnscxXn.nTUfeI1BchcgRG7X0+dzvvPnkuacLOWcfvjgjdm5lmo8be0EM7JObnZw5h3xnuW5TiRc8L.4xGtI1iDe9aNFjKqMoOCvHqfXf4l9GqQqoZgAfPuHbYhUqCPNmQBj5xWCeXga+p5CKbM2cb5GYvthAgfnR0U188MWYmdtYKNrHukiNrxRg97WXq0jzV+8fTZ+8bTuF+nJ+oslFkq5XJSUP8uZ1u9eOU68cbRChJr.+4+1+5YstQefEzfYl0xx145MY5qN11x5B7G0f8FGSFEPOqN5KRWupWqQ77jKd9hzamtvEGjYeOnkditiS5EEouT6p210bq3pK1t5sO17XxsBMhCNEc0lBujBe+toAeeyCiG1oC+bLF8ma0W8IB9QNkcyFV4mlO+BnONU9xY+5G+2+he+C0GmzuPwzYXyGgn.+ev95AekuxXbeamrW1Iw59SJZc+qlGK+P7gs9t6K18p4KDWMAis4Xk0sCOkE.YMJL9tPZh5TlPcoBnbOQnXPOQH2q32qwT9NO1K8.8HoDd0gLJy15HV.dYzYj9YszuT4nexouU+JGk5u9wNZ3RvlCj2L9Rqy7FzWZ8k.JGIFh21zdTHPEqQf++cPLDJOAHILDlRK4+xClmajT9nMKzW8v+A9igYS7YaCylIL+NYO5S8hDelmtsIFK+VJJv4NT8sQqB8jfmIMS5U53zXgFXQP9m44gl662zxpbcV5Fnyx2.cdvMPmUtA5r5MPm2+FnyGbk5fe21GMTJ5qSEABGtsZlEa6zuSt8LV+WzjHKyB

Works alright (for samples shorter than 65500-whatever samples ofc) and I managed to build some other variations for detecting and adjusting according to BPM, beatslicing etc. Pretty fun for lofi and breakbeats but not really for Hans Zimmer stuff lol.

Time Stretching

Block-Based Processing in Third Party Node

rubberband/COMPILING.md at default · breakfastquay/rubberband

13

1.7k

11.8k

103.2k