Please familiarise yourself with the forum, including policy on feature requests, rules & guidelines
Note/sample length for Multi-sampling an Instrument?
NikoBellanti
Madison, WI Posts: 12
Id like to create a sampled instrument from another synth and Im trying to figure out how long I need to sustain the note when sampling each individual pitch for adequate sustain times.
Post edited by NikoBellanti on
Best Answers
-
amiga909
Central EuropePosts: 1,078
Nice topic, I am quite new to multisampling and I am looking for best practices too.
As Icoustik says it comes down to the instrument and what it should do. I did not get far yet with using loop points, it seems to make sense only for sounds that have a repetitive sustain and release phase, for example a basic saw sound with no release time.
Some more points:- Consider it should load fast on the Deluge. It is not like a highend PC with a Kontakt VSTi that can handle GBs of samples. For example huge orchestra sample libraries like Garritan do not work well for me on the Deluge, too slow to load and hogging RAM/CPU quickly, giving me sample dropouts. I went with limiting instrument size in MB to around 25MB. Of course for natural sounds like a real piano this does not work.
- Sample length in s: Length depends on note range (which notes are not sampled, like 0-12, 96-128) and how many samples per octave. I have read around 3, 4 or 6 samples per octave are a good rule.
- Loudness: I found it is better to have them rather loud, it is easier to make a sound quieter than making a quiet sound loud, which can emphasize sampling noise artefacts.
Answers
It depends on the individual sound's decay/ sustain, and the kind of sound - e.g. if its for pads, and you cant make loop points in the samples, you need longer samples, like 10-15 seconds or more.
~ Distinguished Delugate ᕕ( ◎_◎)ᕗ
Thanks guys for the great feedback! Excellent info
I recommend sampling every note. I've tried to sample four notes per octave, I'm very organized about file names etc, but since v3 the Deluge keeps mistaking the pitch of the files and placing them on the wrong note. When I move them to the correct note, the Deluge thinks I want them pitch shifted. It's a pita. So sample every note to avoid the trouble (and the aliasing).
Sampling every note: do you get a problem with file size?
I found limiting file size is crucial for the D. If I load for example about 10 TAL instruments I get note cuts (the D starts to cut samples when reaching its limits), dont know if this is a use case for others. For me this is real problem as my songs tend to max out the D limits (10-15 tracks, 50+ clips). Of course, for other use cases like a song with e.g. just 1 piano multisample, I use the biggest and fattest instrument I have (1GB+).
My projects tend to be smaller than that, so I haven't run into problems with the size of the multisamples yet. But some of the multisamples I created in which the Deluge wrongly pitch shifted my files have been trouble. If I just make a quick multisample and let Deluge do the work it might appear okay but if I edit I find files are often assigned to the wrong root note despite my careful file naming. If you take 27 files and pitch shift most of them it uses a lot of CPU. Single cycle waveforms seem to be the most problematic. So I think there might be issues with both approaches.
What I haven't fully explored yet is if there is a way to turn off the auto pitch detection when loading multisamples, letting the user decide which note is the sample root. That would solve all the issues for me and would allow us to make leaner, more effective multisamples.
I looked for that information in the manual, posted a question about it here, but no joy. Maybe I'll get a chance to revisit the subject over the weekend. I've gotten so busy lately I've given up on creating Deluge multisamples, honestly, but if I find out anything I'll post back in this thread.
@pfrf good points, I concur I would love to have a more in-depth explanation of how the pitch detection works in the manual. I did ask for an option on the D too to import multisample based on filename only.
afaik there 2 ways to override the D pitch detection
interesting you mention single-cycle waveforms. do you use multisampled single-cycle waveforms? never thought about that, I just use them in standard single sample mode.
Thank you for those suggestions. I think I need to get SampleRobot. I'm going to check it out, it might be a big help.
I was experimenting to see what I could accomplish with single cycle waveform multisamples. I wanted to avoid aliasing but things didn't work as intended.
Can you share the XML multisample script please? I would like to make my own multisamples and deluge every time pitch shifted my samples even if they are named alphabeticaly and with note description.
if you know a bit of Bash and if you have a Mac you can roll your own and take this as an example https://github.com/amiga909/deluge-synthstrom-utils/blob/master/multisample/xml.sh
I can recommend using the computation of the XML attributes in the samplesRanges section, got tested a lot and improves a little on how SampleRobot does it.
Unfortunately, I don't know anything about Bash. But thanks for sharing anyway.
Yeah, making a easy to use desktop app out of this is not in my mind. There is Endless WAV and for a desktop app the user still had to get the filenaming right. If Rohan would add the option to load a multi based on filenames only, I would not need a script for myself too.