trying new version,everything seemed fine when altering curves and releases,then i turned up the 'amt' knob (left/top one) and it crashed xt,proper crash,not the domain/pow error thing.
so is that good?
demo mp3 for sifft
-
- KVRAF
- 4760 posts since 26 Apr, 2002 from the bogely factory
-
- KVRAF
- 4760 posts since 26 Apr, 2002 from the bogely factory
tryed again and changing the buffer to a high number crashed xt1
-
obstgegenrechz obstgegenrechz https://www.kvraudio.com/forum/memberlist.php?mode=viewprofile&u=29871
- KVRist
- 335 posts since 17 Jun, 2004 from south germany
nice idea xoxos!!! i love these approaches to making music with samples...
but for now there are quite some things that keep me from using this constantly:
* it crashes reaper 4.261 quite often (last time when changing the sample)
* there are huge CPU-spikes on my system which renders everything unusable for half a second or so... happens often
* the GUI is very small and often i have to go very near my display to actually search for text... sure once i learn the plugin by hard no need for reading
but for its the same for most of your plugins ... also element...
* i find that the selection i can make in the boxes is quite hard to make... for example selecting a filter type feels like a pain for me... would love a rightclick menu...
thanks for all you effort! i really admire all the work you did and thats just heaps!!!
joe
edit: ok now it crashed while changing the time knob...
but for now there are quite some things that keep me from using this constantly:
* it crashes reaper 4.261 quite often (last time when changing the sample)
* there are huge CPU-spikes on my system which renders everything unusable for half a second or so... happens often
* the GUI is very small and often i have to go very near my display to actually search for text... sure once i learn the plugin by hard no need for reading
but for its the same for most of your plugins ... also element...* i find that the selection i can make in the boxes is quite hard to make... for example selecting a filter type feels like a pain for me... would love a rightclick menu...
thanks for all you effort! i really admire all the work you did and thats just heaps!!!
joe
edit: ok now it crashed while changing the time knob...
-
- Banned
- Topic Starter
- 12367 posts since 30 Apr, 2002 from i might peeramid
let me know if there are any more "pow domain" messages, if not, then the envelope is likely solved.
as far as the other crashes, they may be avoidable. when samples are loaded they are fft analysed in full or until the array fills up. unless you have a beefy system, don't load wavs while a sequence is playing, or at least not ones longer than a percussion sample.
changing the buffer also causes the entire sample to be reanalysed. you *could* modulate buffer depth as performance i guess, but it's going to be expensive.
see if anticipating these things stops the crashes and the cpu spikes..
stop the host when a sample is loaded, and trigger it once (that's when the engine will check if the sample is analysed) before you press play. my host occasionally crashes from cpu overload, which i can avoid by not throwing so much at it.
ftr i didn't have any crashes making the 2nd demo.. it started to grind my solitary 1.6g cpu around 5 instances, but knowing when the cpu hits would happen stopped any kind of crash.
as far as the other crashes, they may be avoidable. when samples are loaded they are fft analysed in full or until the array fills up. unless you have a beefy system, don't load wavs while a sequence is playing, or at least not ones longer than a percussion sample.
changing the buffer also causes the entire sample to be reanalysed. you *could* modulate buffer depth as performance i guess, but it's going to be expensive.
see if anticipating these things stops the crashes and the cpu spikes..
stop the host when a sample is loaded, and trigger it once (that's when the engine will check if the sample is analysed) before you press play. my host occasionally crashes from cpu overload, which i can avoid by not throwing so much at it.ftr i didn't have any crashes making the 2nd demo.. it started to grind my solitary 1.6g cpu around 5 instances, but knowing when the cpu hits would happen stopped any kind of crash.
you come and go, you come and go. amitabha neither a follower nor a leader be tagore "where roads are made i lose my way" where there is certainty, consideration is absent.
-
- KVRAF
- 4760 posts since 26 Apr, 2002 from the bogely factory
couldn't get any pow domain errors,so that's good news.
changing buffer gives cpu spikes,but long releases with curve and destination set to pitch gives massive cpu overload.
the finally the amt section set to pitch completely crashes xt when amount is turned up.
(dualcore 3.5ghz,ram:4gig,win7/64)
changing buffer gives cpu spikes,but long releases with curve and destination set to pitch gives massive cpu overload.
the finally the amt section set to pitch completely crashes xt when amount is turned up.
(dualcore 3.5ghz,ram:4gig,win7/64)
-
- KVRist
- 328 posts since 14 Jun, 2004
Hi xoxos,
I can't load a wav with
- 512 kbit/s
- 16 bit
- 2 stereo
- 16 khz
- PCM
- length = 12 sec.
But I can load a wav with
- 1411 kbit/s
- 16 bit
- 2 stereo
- 44 khz
- PCM
- length = 1/2 sec.
I rendered (with reaper) the first one with different khz values and different bit rates with no success. What is wrong?
I can't load a wav with
- 512 kbit/s
- 16 bit
- 2 stereo
- 16 khz
- PCM
- length = 12 sec.
But I can load a wav with
- 1411 kbit/s
- 16 bit
- 2 stereo
- 44 khz
- PCM
- length = 1/2 sec.
I rendered (with reaper) the first one with different khz values and different bit rates with no success. What is wrong?
-
- Banned
- Topic Starter
- 12367 posts since 30 Apr, 2002 from i might peeramid
akira - the wav loading code is rudimentary.. it loads all the samples in my folders but it's not prepared to deal with files with unusual headers, headers with extra data i them. it checks 8 times sequentially for the 'data' block and if it can't find it, it doesn't load, so you will find samples here and there that it won't load.
spacedad - cpu goes up with pitch.. the fft buffers are read at varying rates by pitch.. so eg. if you're using a 1024 buffer and playing 10 octaves high, you're going through a lot of buffers... 2^10 = 1024, so you're computing a new buffer on every sample at +10 octaves (well two actually since buffers are overlapped). the frequency shifting isn't nearly as expensive.
spacedad - cpu goes up with pitch.. the fft buffers are read at varying rates by pitch.. so eg. if you're using a 1024 buffer and playing 10 octaves high, you're going through a lot of buffers... 2^10 = 1024, so you're computing a new buffer on every sample at +10 octaves
(well two actually since buffers are overlapped). the frequency shifting isn't nearly as expensive.you come and go, you come and go. amitabha neither a follower nor a leader be tagore "where roads are made i lose my way" where there is certainty, consideration is absent.
-
- Banned
- Topic Starter
- 12367 posts since 30 Apr, 2002 from i might peeramid
i either save the wav with another editor (sometimes by copying and pasting into a blank file), or load the wav in a host (or similar app) and record it that way. and cross your fingers
you come and go, you come and go. amitabha neither a follower nor a leader be tagore "where roads are made i lose my way" where there is certainty, consideration is absent.
-
- KVRian
- 1078 posts since 28 May, 2003 from world
latest sifft (either the normal or 16 voices version) doesn't produce any sound in Renoise 2.8.1 anymore after loading a .wav into it. , can't seem to change the buffer size either.
-
- KVRist
- 328 posts since 14 Jun, 2004
Hi xoxos,
Wavoaur did it!
Now some feedback:
1. A wav with 3:51 minutes was cut down from Sifft to 11 sec. ....is there a limited length for wavs loading into Sifft?
2. With a certain value of the OCT knob I get an unprocessed (not bypassed) signal. If I then increase the time knob, I get a wonderful timestretched signal.
But when I decrease the time knob again to zero, I don't get the unprocessed signal again...it sounds, like there is a fast LFO on the original signal (maybe the stretch-algorythm is still working). That's a pity, cause it's impossible to come from a weird bizarre sound back to the normal, natural voice.
3. When I changed the values in "Interpolate" and/or "Buffer" and played a note on the keyboard Reaper instantly crashes.
4.
5. Always liked people beside of the mainstream.
P.S.: I'm not sure yet, but I think we need a more detailed manual
Wavoaur did it!
Now some feedback:
1. A wav with 3:51 minutes was cut down from Sifft to 11 sec. ....is there a limited length for wavs loading into Sifft?
2. With a certain value of the OCT knob I get an unprocessed (not bypassed) signal. If I then increase the time knob, I get a wonderful timestretched signal.
But when I decrease the time knob again to zero, I don't get the unprocessed signal again...it sounds, like there is a fast LFO on the original signal (maybe the stretch-algorythm is still working). That's a pity, cause it's impossible to come from a weird bizarre sound back to the normal, natural voice.
3. When I changed the values in "Interpolate" and/or "Buffer" and played a note on the keyboard Reaper instantly crashes.
4.
I support this whole-hearted. I think, that some of the developers are nerds, who can't imagine, that there are older people, or people with weak eyes, or people with a big display and a very high resolution, where you need a microscope to read some GUI's. Xoxos...couldn't you make it bigger and increase the contrasts?obstgegenrechz wrote:* the GUI is very small and often i have to go very near my display to actually search for text... sure once i learn the plugin by hard no need for reading
5.
I support this whole-hearted too!obstgegenrechz wrote: thanks for all you effort! i really admire all the work you did and thats just heaps!!!
joe
Always liked people beside of the mainstream.P.S.: I'm not sure yet, but I think we need a more detailed manual
-
- KVRAF
- 4760 posts since 26 Apr, 2002 from the bogely factory
...or you could decrease your resolution.akira wrote:...people with a big display and a very high resolution, where you need a microscope to read some GUI's. Xoxos...couldn't you make it bigger and increase the contrasts?
i don't know,people these days have such high res that they can't even see a single pixel!
-
- KVRAF
- 1618 posts since 15 Aug, 2001 from montreal, canada
Really enjoying this little gem xoxos. I do manage to crash it each in audiomulch
though. If I load a sample and set the buffer to 16384 or 32768, it crashes as soon as I play it.
though. If I load a sample and set the buffer to 16384 or 32768, it crashes as soon as I play it.
-
- Banned
- Topic Starter
- 12367 posts since 30 Apr, 2002 from i might peeramid
i'm thinking these crashes are performance based, that too much processing is being thrown at the system at once. i'm using eXT2 and if i load samples for six instances at once, my host freezes for about three seconds while the computations are performed, but doesn't crash. i guess i'm lucky as i haven't been able to make it crash :p
next time i code an fft based sampler, it will read the analysis frame in real time, which will cost cpu but will allow for longer samples and flexibility. there's no time limit, it's about data that can fit in an array.
without explaining everything that's involved, i'd recommend using short samples and buffer lengths under the sample length. if you select an 8000 sample hit and set it to 16384, it *should* start playing once you set the buffer back to 4096 or under.
don't try to punish it.. i write synthesis code.. i know *nothing* about how you have to baby operating systems with high performance, so that's why it's bottlenecking processors. i can't stop it from crashing if its trying to do too much at once.
as far as i know, there is one case where the code could cause a crash (you don't want a technical explanation), which is related to loading large samples (let's say 30 seconds or longer).
akira - what you're hearing is "vertical phase decorrelation.." "vertical phase correlation" is something sifft doesn't do harmor et al use methods i've yet to find out about to create the impression of v.p.c. but it's intrinsic to this method of timestretching (check the manual pics on page 2.. that's why it sounds "original" if you trigger it from the beginning, but phasey once you do anything to it..)
ftr, paulstretch is about at the same technology level (unless he updated). i remember reading that his solution was to randomise the phase, which is why paulstretch sounds ~reverby on long stretches - that's the same behaviour you're hearing in sifft.. if you were analysing a sample with fft and laid the frames out in series, "horizontal" correlation would be retaining the phase relationship from one frame to the next.. "vertical" correlation is retaining the phase relationship between frequencies in a single frame...
horizontal is necessary to produce the correct pitch, vertical to produce the correct transients/reassemble the data with the original phase correlation (or you get diffused reverb). the advanced methods use tricks like maintaining vertical correlation around spectral peaks and processing them in groups. i can't make that yet.
i kinda think of it as the 8 bit edition of something like harmor.. it doesn't produce indistinguishable resynthesis, but it sounds kinda techno when you do stuff with it.
next time i code an fft based sampler, it will read the analysis frame in real time, which will cost cpu but will allow for longer samples and flexibility. there's no time limit, it's about data that can fit in an array.
without explaining everything that's involved, i'd recommend using short samples and buffer lengths under the sample length. if you select an 8000 sample hit and set it to 16384, it *should* start playing once you set the buffer back to 4096 or under.
don't try to punish it.. i write synthesis code.. i know *nothing* about how you have to baby operating systems with high performance, so that's why it's bottlenecking processors. i can't stop it from crashing if its trying to do too much at once.
as far as i know, there is one case where the code could cause a crash (you don't want a technical explanation), which is related to loading large samples (let's say 30 seconds or longer).
akira - what you're hearing is "vertical phase decorrelation.." "vertical phase correlation" is something sifft doesn't do
harmor et al use methods i've yet to find out about to create the impression of v.p.c. but it's intrinsic to this method of timestretching (check the manual pics on page 2.. that's why it sounds "original" if you trigger it from the beginning, but phasey once you do anything to it..)ftr, paulstretch is about at the same technology level (unless he updated). i remember reading that his solution was to randomise the phase, which is why paulstretch sounds ~reverby on long stretches - that's the same behaviour you're hearing in sifft.. if you were analysing a sample with fft and laid the frames out in series, "horizontal" correlation would be retaining the phase relationship from one frame to the next.. "vertical" correlation is retaining the phase relationship between frequencies in a single frame...
horizontal is necessary to produce the correct pitch, vertical to produce the correct transients/reassemble the data with the original phase correlation (or you get diffused reverb). the advanced methods use tricks like maintaining vertical correlation around spectral peaks and processing them in groups. i can't make that yet.
i kinda think of it as the 8 bit edition of something like harmor.. it doesn't produce indistinguishable resynthesis, but it sounds kinda techno when you do stuff with it.
you come and go, you come and go. amitabha neither a follower nor a leader be tagore "where roads are made i lose my way" where there is certainty, consideration is absent.
-
- KVRAF
- 4760 posts since 26 Apr, 2002 from the bogely factory
man,i just load 1 short kick sample and turn up the pitch thingies and boom.