I'm new here. Apologies if this is in the wrong place.

Hello All,

I am working on a performance piece in which a live input needs to be timestretched.

I understand timestretching a live input may be impossible, however I was hoping a granular delay approach may work.

To help explain - if you imagine a delay box set to a very short delay time like 2ms, and a long decay time like 2 seconds, you would get a ringy layered delay effect. This is what I want to do but in a clear timestretched manner, but still layer the sounds as they come...

Does this make sense.

Can any one point me to a patch or technique ?

Thanks
Twogan

GC - Thanks so much for the detailed reply.

I do understand your description. I like your runner metaphor.

I am again worried that I am trying to do the impossible.

I need to be able to constantly stretch a stream of incoming audio with no gaps or latency. I ultimately want to use this as an overall effect applied to the stereo mix of a live band playing. I want them to start playing vvvvvvveeeeeerrrrrryyyyyy sssssllllooowwwllllyyy wwwwiiiitttthhhh aaaaaa llllloooonnnnnggg sssstttttrrretchhh,,,,,,, ttthheeennn ggrraadduuaalllyy ddeeccrreeaassee tthhee ssttrreettcchh aammoouunntt aass tthey sspeed uup, uuntil aafter ssome ttime tthey aare jjust playing clean and tight.

To reiterate - the band would start by physically playing their instruments slowly, whilst there is a long stretch amount applied. Then as they start to physically play faster, we lower the stretch amount in accordance..

EG, if the source was a snare drum being hit once every 4 seconds, and each resulting hit sound was stretched to 3 seconds long you would end up with a nice one second gap. With a constant sound as an input (say a bass tone) we would end up with a massive noisy pile up of audio (which would be fine by me). If the snare hit and the bass tone were played together you would end up with a big noisey mess of bass tone over a nice 3 second long snare hiss...

Do you see?

To follow your runner metaphor I would need the runner to be constantly spawning hundreds of new runners who would slow down and fade away I suppose...

Its almost like to get rid of the delay in the patch below I would need very many versions of the patch taking an input once after each other in turn, all delayed by the same amount as the grain size?

i.e - So instead of 'turning off' the input whilst the stretch is done, its constantly 'on' and the stretched output is layered over itself...

Does this make any sense?

Its almost like I need the effect to function in the way a reverb does, rather than a time based thing...

Sorry, hope that makes sense. I am making my own head hurt trying to explain this

twogan

OK so I have been experimenting with the timestretch RT.pd patch linked above.

I have found that it is exactly what I need however - it has some issues.

I find that there is a big variation in the 'latency'. When I first use the software its not too bad. but if I move the sliders it becomes very very long - up to about 8 seconds... So not really usable...

Can anyone see why this might be the case? Perhaps it is just on my set up...

Thanks
Twogan...

Great. Thanks for the clarification GC

Many Thanks for this.

I cannot see the adc~ or dac~ so how do I route my mic input to this ?

Very Sorry for my naivety..

T

This all sounds very hopeful indeed. I'm glad I am not asking the impossible.

Many thanks indeed for the ideas and advice.

I am a beginner with PD and I am afraid a lot of this is going over my head ?

How long do you think it would take me as a beginner to get up to the level where I can make this patch myself ?

Are there any existing patches available that I can look at / modify ?

Is PD the best way to go to achieve what I need to do, in view of my inexperience ( I have used Processing, PD and arduino etc... before but always quite simple programs for very specific projects, usually involving sensor input, not audio processing ) ?

Many thanks in advance for your time on these questions

Twogan

Thanks you for these replies.

It is a tricky concept to describe and I too am having some difficulty understanding these replies.

They both sound promising.

Just to reiterate:

I want to timestretch the live input from a microphone in real time. But obviously as the software cannot wait for the entire 60 minute performance to end before outputting the stretched version, I need this to happen immediately, so the incoming timestretched sounds are layered on top of each other

So sound would go from:

Boom boom boom

To:

Bbbbooooooooooooobbbboommmmoooobbbbooommmmooooooooooommmm

Is this understandable ? Would this be the result of the techniques in the replies?

Many Thanks
Twogan