Home | ER-101 | ER-102 | ER-301 | Wiki | Contact

Techniques for Real-time Data-driven Synthesis

Love the video. That’s a really amazing use-case you have put together there… taking any mono source and making it poly! Brilliant!

1 Like

yeah - for it to be perfect, the punch would have to be at the start cycle of a wav, and the punch out would have to be exact as well when the wav finishes.

2 Likes

This is right - I wouldn’t be too surprised if you could sync the two and actually achieve this!!

It is a super cool idea!!!

2 Likes

I have been trying this actually - using a tap tempo unit with it’s tap modulated by the source wave form, and then using the tap tempo to reset the sample loopers and players. No love though.

Have also tried the pitch shifting delay instead of sample players at 100% wet and using a V/O to control the speed, though this does not seem to be set up for 1 V/O (and I’m not sure it would produce a good result if it were).

Nothing’s really working 100% seamlessly. Probably the best compromise for now if you wanted to do this is to do like I did in the video, set a moderate buffer length (maybe 0.5-1s) to balance between masking the artifacts and minimizing the lag time of adjustment of the external OSC, and when you get something you like, record the source wave use a DAW to trim it to an exact cycle length.

I’ve done a few experiments here, and like Neil mentioned, the purer the waveform, the more noticeable the reset event is. Using it with some complex phase modulated signals (which really was the goal - no need to clone a sine wave here), its less pronounced. I did actually get some results that I was having fun playing on the keyboard. Though if I were to use them in a recording I’d have definitely trimmed them to a cycle length and used them directly in a sample player.

I will probably move on for a while, and maybe come back later. Lots more fun things to explore :slight_smile:

I guess this is what I would have suggested to try first. It’s a workaround but for now 1V/oct control is possible via

1V/oct --> Sine Osc(1V/oct) --> Period-o-meter --> Pitch-shifting Delay(delay)

EDIT: Actually scratch this. I had karplus on my mind when I wrote it. The Pitch Shifting Delay has a speed input (linear FM) and at the moment there is no way to convert 1V/oct to a speed in the ER-301. So I should really provide a 1V/oct input on the Pitch Shifting Delay.

Another alternative is the Manual Grains unit sharing buffers with a Sample Recorder. You might need some creative control of the grain position to advoid the discontinuity created by the record head in the looper.

1 Like

So one of my personal goals with the ER-301 is to support the process of composing with (long) field recordings. I am especially interested in removing the need to pre-process or label or edit long field recordings before they can be used effectively. This encompasses a whole bag of techniques but one relevant to this discussion is the ability to move a pointer around in a long field recording and use nearby(*) audio to generate tuned amplitude-normalized oscillators. I guess you could think of it as real-time wavetable generation from an audio buffer whose contents are constantly changing. Whether it actually uses wavetable techniques under the hood or something closer to granularization is something that I am still in the process of evaluating.

(*) I use the term nearby in the generic sense. One naturally thinks that nearby means a local neighborhood in the sample buffer. However, you can also group the audio so that parts of the sample buffer that “sound similar” are also considered nearby. This way you have additional degrees of freedom in which to move the pointer: not just later or earlier in the buffer but also to another temporally unrelated but similar sounding section of the buffer. Of course, the real challenge will be creating a total ordering on top of this audio similarity metric so that browsing it makes as much as sense to the user as scrubbing a play head over a time line. This is the part I’m not sure I will succeed at.

21 Likes

:open_mouth: I’m stunned…this sounds amazing…definitely pursue it!

this is awesome and i’ve been inching towards it mentally so it’s good to see. :slight_smile:

@Joe I would be more inclined to use an external source for sync - it could just be the trigger for each note sent to both the sync input on your oscillator and the ER-301 - that way the oscillator and the recorder may just work?

It would definitely be something I would try anyway :slight_smile:

@odevices Joe is a very clever chap and has endless ideas that are both inspirational and challenging - we have worked together on a variety of projects - I really must get my act together and finish off the big one we did earlier this year and release it into the wild :wink:

1 Like

Thanks for splitting this topic - it was starting to feel very misplaced to me too. :slight_smile:

That’s a very nice compliment. Thank you, @anon83620728! I tend to think of myself more as “even a blind squirrel gets a nut once in a while” and I am a lucky squirrel. :stuck_out_tongue:

This whole idea sounds amazing. I think I sort of loosely get what you’re talking about with total ordering applied to a field sample. It sounds very mathy. But I believe in you, Brian-san! Will be interested to watch this unfold and hopefully be part of it.

Do you mean the gate signal from, in this case, my keyboard?

When I tried using the pitch shifting delay before, I was definitely hearing something strange, so I thought I’d explore it a little further and shot another video. It seems to be affecting the amplitude of the input. This may be user error, but if not then it the pitch shifting delay might need a few more tweaks before it could be used for this purpose. As I think about it though, if it had a V/O control and didn’t have the behavior shown below, it actually is probably more ideal for the live cloning idea.

EDIT: Geesh, wherever I say “variable delay” in the video, just mentally replace that with “pitch shifting delay”. I’m off to the coffee pot now.

Sure, why not?

I suppose it will depend on how your oscillator implements sync, but a gate should work, if not a gate to trigger before the sync?

That would be phase cancellation where two grains are overlapping but with different phases. It’s especially obvious with simple waveforms. One possibility is to phase lock the grain onsets but then you will probably hear some amplitude modulation from the grain envelopes that are no longer exactly lined up. Actually, I’m not aware of a granular method that has no artifacts.

I can tell a difference in the quality with the pitch shifting delay in 0.2.11. Geesh, I am so close to cloning an external oscillator for my evil purposes, I can taste it. :speak_no_evil:

5 Likes
3 Likes

Here’s how I feel: I purchased an ER-301, and in addition to a new module I’ve been given a key to a room full of geniuses.

14 Likes

ok i have a weird issue with a click i hear and i have absolutely no idea why it is there…
simple setup, some phase modulated sine wave drones go to main monitor and a copy of it into erbeverb and back into the 301 where it gets recorded into a looper with full dub and full wet and a 20ms fade, next to the looper i have a sample player with the same buffer as the looper so i don’t hear the looper but the pitched down sample player, again with a rather large fade. still, i hear a click and i don’t know why, i watched both the looper and the sample players scope to see if it’s only at the end of the recording but it doesn’t match. do i misunderstand this concept of looper/sample player? the waveform on the player’s scope does not update the recording but the buffer clearly does update, i wonder if it’s somehow connected to this shared buffer…?

Change the fade on the looper to 0ms.

Scratch that. I thought you were listening to the output of the looper. The click is the looper’s record head impacting with sample player’s play head (which is going slower).

I have to say I have noticed the same “click” at the end of the loop and thought it might have been something I was doing. You can visually see the spike at the end of the waveform. I’m also using a trigger to control the reset on the looper to keep it in sync.

hm, i don’t fully understand. the click is the sample player updating? is there any way to avoid this or is this normal when the sample player isn’t at normal speed?