How to achieve realistic Legato / Slide / Portamentos?

Happy New 2018 everyone

As far as I’m concerned, Renoise doesn’t support “multisample-legato”. Correct?

I’m trying to make realistic legatos / glides / portamentos with Renoise and I’m not sure how.

What are your workarounds for this?

If you’re unsure what I’m talking about, take a look at this simple example:

As you can see, this is just a glide with the GXX-pattern command from C-6 to G-6.

Let’s say you have key-mapped one sample for C-6 (up to F#6), another sample for G-6, and so on… You know the drill.

So, one might expect the G-6 sample to play at line 04, but drumroll it doesn’t. The C-6 sample is still going but simply up-pitched to G-6 (while ignoring the G-6 sample altogether).

Granted, sometimes you want this particular effect; the “smurf-effect”. But for more realistic sounding instruments; voices, orchestral, etc, you certainly don’t want that - unless for experimental purposes. Heck, even some rendered synth-leads can sound a bit strained. The same behaviour seems to be the case with similar pattern-commands such as DXX / UXX.

I suppose the feature would only crossfade between the samples, and/or some other smart solution. Still that may not be as realistic as a true legato (live-recorded legato at source), but probably a lot better result presumed the samples are tuned correctly etc.

Whatever the solution may be, it would be a nifty option to the standard behaviour?

The only workaround I know of is to not program with the Renoise-sampler but instead use a plug-in which supports the feature, and program accordingly (MIDI or some other sampler). Yet, us around here, love the intuitive way to program with Renoise commands.

So, again, with the current version of Renoise, can a more smooth “non-smurfy” note-transition be programmed with samples?

I dont think there is any way around it.

Renoise always just pitches up or down the sample that is assigned to the key which has the glide, pitch up or pitch down commands applied to it.

Its fine for single cycle looped samples.

For other samples in which the looping section is longer it can sound a bit stange if you slide more than one octave. But most slides in a piece of music will usually be less than an octave.

I would not put a bet on this to be implemented somehow. Maybe you could already “fake” it, but it would require serious hackery and kill any fast instrument generation workflow.

For voice you might have better success with some autotune like plugin, or other dedicated pitch shifters, using them to redefine the pitch of a voice track, it will be capable of handling the formants to keep them intact and not become grunts or mickey mouse.

Maybe you also might find greater joy in some kind of spectral resynthesis vst plugin, or granular synthesis which is capable of similar warpings but will sound more foggy or gritty.

You know, there are certain features that are similar, but probably won’t come into renoise all too soon because of extra implications. For example velocity crossfading is similar to pitch/portamento crossfading that you suggest. To make them sound any good, you will have to have the samples used close to 100% phase aligned, which will require vast amount of care while building an instrument from samples. Else the transitions have a very high chance of sounding like cack, for any instrument that uses samples that are more complex than single cycle waves of the exactly same tuning.

I like using only the renoise sampler, but I do much sound design to the final tone, so working from very simple wavs into shitloads of dsp that will be running live…the dsp can then define formants that aren’t shifted with the pitch like they would if I resampled a single note and did portamentos on the resulting sample.

But most slides in a piece of music will usually be less than an octave.

Yup, “usually” is the correct term. It’s funny to pitch samples wildly, but like you say, it really depends on the source material.

For example velocity crossfading is similar to pitch/portamento crossfading that you suggest. To make them sound any good, you will have to have the samples used close to 100% phase aligned, which will require vast amount of care while building an instrument from samples. Else the transitions have a very high chance of sounding like cack, for any instrument that uses samples that are more complex than single cycle waves of the exactly same tuning.

Again, yup. But this time, it’s something which is very likely to happen. Muffled, weak sounding tones could easily become the result of cross-fading.

Some serious DSP magic would need to be implemented to avoid this from happening in realtime (an interesting challenge though)

Thanks guys for contributing to the topic.

Yup, “usually” is the correct term. It’s funny to pitch samples wildly, but like you say, it really depends on the source material.

Actually, I do pitch wildly. My synth-solos are on crack!

Sure, sometimes the current behaviour does some interesting artefacts which unintentionally made an interesting outcome. But sometimes when the pitch-slides are “semi-cracky”, it just sound strained. I’m talking about any type of sound-source now.

Out of curiosity, have the Renoise Team debated / considered the topic? I’m sure you’re aware of that there are scripts for Kontakt which has been made to do this kind of things by volunteers. Can a Renoise Tool be written in similar fashion? I would economically support such a project, and with the disclaimer of un-satisfied results.

Furthermore, all modern sample-libraries have some kind of work-around / options to this if true legato [legatos recorded at source] isn’t supported. At least, I think the topic should be considered for further discussion.

For voice you might have better success with some autotune like plugin, or other dedicated pitch shifters, using them to redefine the pitch of a voice track, it will be capable of handling the formants to keep them intact and not become grunts or mickey mouse.

Yep. It’s a good workaround when working with voice-type material sometimes.Maybe a similar option to pitch-bending, like how Melodyne works, could be implemented directly in the Renoise sampler? Like a formants-option in the sampler, which would work with the current G/U/D commands (related to pitch).

In case someone else further on is interested regarding the Melodyne-workaround, which I’ve tried, and reading the topic:

I would not put a bet on this to be implemented somehow. Maybe you could already “fake” it, but it would require serious hackery and kill any fast instrument generation workflow.

—>

I like using only the renoise sampler, but I do much sound design to the final tone, so working from very simple wavs into shitloads of dsp that will be running live…the dsp can then define formants that aren’t shifted with the pitch like they would if I resampled a single note and did portamentos on the resulting sample.

Would you care to elaborate?

Maybe you also might find greater joy in some kind of spectral resynthesis vst plugin, or granular synthesis which is capable of similar warpings but will sound more foggy or gritty.

Then again, I won’t be able to use the Renoise commands?

The workaround for now is to find the most suitable sample-library, mostly Kontakt-related although I’m not entirely sold on Kontakt. It’s not anywhere near the simpleness of Renoise. Admittedly, it may just be a matter of personal preference but each Kontakt-library has it’s own conditions to be learned from a user pow. So I might just as well use any plug-in then.

You know, there are certain features that are similar, but probably won’t come into renoise all too soon because of extra implications. For example velocity crossfading is similar to pitch/portamento crossfading that you suggest. To make them sound any good, you will have to have the samples used close to 100% phase aligned, which will require vast amount of care while building an instrument from samples. Else the transitions have a very high chance of sounding like cack, for any instrument that uses samples that are more complex than single cycle waves of the exactly same tuning.

I know, but let’s say the samples are correct with no phase-problems. So back to my example and with this hypothetical feature; I’m thinking the G-6 note should not retrigger the modulation (if any such additions has been set like a random LFO etc) but rather continue whatever the C-6 note triggered although the sample of G-6 is different.

I respect the difficulties to implement the feature. But most modern releases of samplers and sample-libraries have taken this issue seriously, so it will be a matter of interest and priorities like anything else. It couldn’t be stressed enough that my suggestion should only be an option of reasons already mentioned.

It is not really difficult to implement in a naive way - just crossfading 2 (or more over a longer run) samples. Difficult is to prepare samples that will have to blend perfectly. If they don’t match, they will sound a bit like through a blanket or dull reverb while they are blended. If you wanted any samples to morph this way without extra work matching them, you would need quite advanced tech, which automatically prepares samples spectrally for you to be morphed.

Elaborate the technique of keeping formants live processed? Well, say you have a saw, and a filter. if you sample the saw through the filter, and then do a pitch glide, the filter will shift with the pitch, or rather…it will become more compressed (up, mickey mouse) sounding or more “wide” (down, grunt sgnt). But if you don’t resample the filter, and keep it running live, the filter will stay at the position where it is when pitch is gliding, and it even is movable independently. So if you have some vowel like synthetic sound, you could stay away from using a sample with that charakteristics, but just feed some raw tone through a vowel filter, and the formants are independent of the pitch you play the tone at. This is what I meant by keeping processing live, don’t reample the filters, and you will evade the mickey mouse feeling. I build quite complex instruments around this techniques, and sometimes I even transpose filters the way they would move when they were resampled - if the desired tone requires that action, and I do not want to resample a filter earlier in chain. Maybe you’re a bit disappointed now, because it is so trivial and it would mean to ditch the ready made samples and learn how to design the tones yourself. Ofc this isn’t always practical and designing complex instruments means hours over hours of work, but well…this way you get timbres that won’t shift with pitch, or even shift, for example, into the opposite direction which sounds very cool on upwards bends.

It is not really difficult to implement in a naive way - just crossfading 2 (or more over a longer run) samples. Difficult is to prepare samples that will have to blend perfectly. If they don’t match, they will sound a bit like through a blanket or dull reverb while they are blended. If you wanted any samples to morph this way without extra work matching them, you would need quite advanced tech, which automatically prepares samples spectrally for you to be morphed.

Elaborate the technique of keeping formants live processed? Well, say you have a saw, and a filter. if you sample the saw through the filter, and then do a pitch glide, the filter will shift with the pitch, or rather…it will become more compressed (up, mickey mouse) sounding or more “wide” (down, grunt sgnt). But if you don’t resample the filter, and keep it running live, the filter will stay at the position where it is when pitch is gliding, and it even is movable independently. So if you have some vowel like synthetic sound, you could stay away from using a sample with that charakteristics, but just feed some raw tone through a vowel filter, and the formants are independent of the pitch you play the tone at. This is what I meant by keeping processing live, don’t reample the filters, and you will evade the mickey mouse feeling. I build quite complex instruments around this techniques, and sometimes I even transpose filters the way they would move when they were resampled - if the desired tone requires that action, and I do not want to resample a filter earlier in chain. Maybe you’re a bit disappointed now, because it is so trivial and it would mean to ditch the ready made samples and learn how to design the tones yourself. Ofc this isn’t always practical and designing complex instruments means hours over hours of work, but well…this way you get timbres that won’t shift with pitch, or even shift, for example, into the opposite direction which sounds very cool on upwards bends.

Hi, sorry for a bit late reply. But among other things, I had to come back to the post 2-3 times until I understood what in the world you were talking about. :lol:

But I think I understand now. Either crossfade the samples or build your own instruments using the internal synthesis of Renoise, which both of course is very time-consuming. But thank you nonetheless for the in-depth explanations. Very nice of you to share.

As a matter of coincidence I have lately found the synthesis-process (within Renoise) quite interesting which I have already experimented a bit with, and hopefully will continue on doing when time allows. It certainly feels like the more “proper” workaround of the two mentioned, and there are of course a lot of other advantages with doing your own instruments in general.

I have two additional questions if you don’t mind.

There are two ways of pitch-bending in renoise, one by sample-commands and one with the DSP’s, correct? Do you have any thoughts of the advantages / disadvantages of those?

Do you mainly use Renoise for building your own instrumens, and / or do you use plug-in synths? For instance, I love the sound of FM8, and it does glide “properly” (because of the internal synthesis I suppose). But as earlier stated, I find the Renoise commands a lot more intuitive than various MIDI-commands etc, which will pay off no matter if the DSP’s are of lower or higher quality than plug-ins. Also, speaking of time, going through the learning-process all over again with each and every plug-in is maybe not preferable.