Hi, there are few times when to get by ear all notes in a fast solo I slow down / time stretch pieces , and mostly works, but in extreme fast cases for percussive sounds like piano that means that the attack time is smeared so much (could be 0.5 secs) that is difficult to get the right timing.

So IF I export the slowed audio, what can I do to restore at least a bit of the sharp attack?
Transient shapers?
BTW if also decay could be shortened a bit it won't be that bad.

If possible I would prefer standalone sw, but in case it's not possible also daws / vsts.
Windows only.

 

You can use Transient Shapers, but you are wanting a standalone solution so a Spectral Editor is probably where you want to go with it. So yeah, RX or Spectralayers. Still, if you time stretch something past a certain point, fixing the smear and attacks is only going to get you so far. You might stick something like SPL Transient Designer, or something like that, and maybe even Newfangled Punctuate at the end of a chain you will probably have to build each time, or close enough with reconfiguration of it.

Would Zplane DeCoda help you not break it in the first place at all? That would be a workflow thing.
Just a couple of ideas. More details about the workflow might help someone who does this too. (I do not).

 

Any stretch tool at that level of slow down (even 1/4 or less) I tried work the same way, that is not a faulty behaviour, I mean if I input a snare and stretch it down a lot it becomes puffy, for a piano much more than that.

So I'll try some tool as you explained, but I don't expect great results.

 

Try it with TBTech Trinity Shaper. Enable multiband mode and then try to find the transients. There's a lot of fine tuning controls, you might have good luck with it.

 

Any stretch algo you use is going to affect the transient, depending on how much you stretch it and the algo itself. I suggested Decoda because to me, your first post sounds like you are trying to learn songs, to increase readability when you say "by ear", which is what is is really for. Instead of repairing the damage, you could also try lessening it, but you did not specify the workflow. Are you using Elastiquepitch, your DAW, Serato Sample, etc? Sometimes the best way to fix damage is to reduce it or eliminate it happening in the first place. Since you are on PC, are you using old Protools ever? R2R released Pitch N Time from Serato, but it is AAX only iirc.

 

Last edited: Jun 3, 2026 at 10:09 PM

I don't think that with a 0.5 second stretching the there is much you can do to sharpen the attack, but one can more easily in my mind shorten the length of the note to help maintain the imposed slower bpm. Someone else might recommend a plugin (envelope follower?) for that as my mind is fairly mush around about now. You can perhaps also after the stretch add a fresh and in time click track in the now slowed track to help orient yourself with your playing along with it, something that is transient rich.

 

Last edited: Jun 4, 2026 at 2:58 AM

^ For shortening hits, you can use an envelope, a transient designer, a gate, or even a de-reverb plugin. They all reduce the apparent length of the sound in different ways. With transient designers, controls like Decay, Sustain, or Release.. whatever the plugin actually calls them, usually affect the tail of the sound, making the hit shorter and tighter. It's a good trick to use on fast stuff where you want space, but also for cleaning up the "midi spam" if you convert audio to midi data.

But a .5 second stretch wouldn't pose a problem, he is actually talking about a .5 second smear. It sounds extreme to be that much, but he is slowing down a lot if that is a real time measurement.

 

I use Transcribe! for learning solos, it will even show peaks with note names. It's very good at retaining the audio quality when slowed down and it has other tricks such as changing octaves. Also easy to add placement tabs along the timeline and loop sections.

 

ASAP I'll try all the suggested tools, though I don't expect nothing awesome.

For ensemble tracks, in tools like decoda and similars the piano roll is most of times cluttered and garbled, it's not much useful, though the time stretch function is really precious.

With solo instruments it's much better but I found Aurally Sound Prism the best among them.
I played a piano piece, then fed in, the midi result was astonishing, even pedal was retained in let's say 98%.

 

Your best bet is using an audio to MIDI solution.
so you can slow it as you want.

But of course, the software do the transcription itself.
If you combine it with track demixer when there are different instruments, it can works even better.

 

I have used Prism quite a bit and in dense mixes or strongly separated material you still tend to get what I call MIDI spam, meaning lots of small fragmented note events that are not played notes in the audio file. It does not outperform Melodyne Studio for polyphonic detection. It is not magic.

A key point that often gets overlooked is that polyphonic transcription depends heavily on transient definition. The clearer the separation between note start, the more stable the pitch and note extraction will get . When attacks are smeared or blurred by stretching or separation, transcription tools lose reliable extraction and start producing fragmented or unstable results. Sometimes I will set a very fast Midi arpeggiator on them and bounce it again, and it works a decent amount of the time.

Like @Lois Lane already mentioned, SPL Transient Designer can help shape the notes onset clarity. More advanced processing in tools like Accentize, RX, or SpectraLayers can improve separation and reduce masking, but it also comes with a more manual workflow.

Also, when people talk about ensemble tracks, it is important to mention whether they are being stem separated first or not. If you are using UVR5, the model choice and build version matter a lot. Different Roformer builds and models can produce very different transient behavior and separation artifacts, and that directly affects how usable the material is for transcription or as midi note data to trigger another synth/drum machine.

Results vary less because of the transcription tool and more because of how clean the source material is before it enters the transcription stage. If it really is an issue of your timestretching everything first, I'd really consider PT and Pitch N Time from Serato. It's an old tool and algo, but it's like 700 bucks still for a reason.

 

I was thinking to do that, just for selected troubled snippets.
But I don't need to midi them, because then I have to import them in daw, assign an instrument etc. BTW imported midi scores without editing are a nightmare to read.
I just need to play along the exported audio measures/snippets in order to memorize them, then I increase speed gradually up to original tempo.

Prism is designed for solo instruments, though can be polyphonic.

 

I've only used it on single stems, and a single solo instrument can still be polyphonic. If I want the individual notes of polyphonic chords, I always just go straight to Melodyne Studio. Melodyne is considered the best for polyphonic note detection because it is it reliable in real production workflows. Their own "lite" versions do not even do it.

Have you ever tried RipX DAW for this? When all the stem separation tools first hit the market, the results from their earlier products like Hit N' Mix Infinity, and then DeepRemix/DeepAudio were incredible compared to everything else and it was very fast, even on old systems. There was a very short period of time when they were the go-to for stem separations, before AI models caught up very quickly.

I just started checking out V8 of the actual DAW that emerged from the early products, since I just got an M4 Mac to try it with, and when you first start to import a track for ripping, you get this menu of options, check the bottom right one:



Ignore the 10 mins estimated time, that is for the 8 minute long track I just selected to rip. It will be usually be a lot faster rip time than most models, unless you use cloud services like MVSEP. The Wavs only stem save shows 2 minutes for the same track. It was not as high quality separations as certain models like Roformer or even HtDemucs_FT, but maybe it doesnt matter to you ;]

 

I'll try but I don't think it will solve the slow down smearing.

However, the issue I posted is just for very few tracks, for most of them I stretch between 75% and 50% and I can cope with that.

 

I just got around to trying it on the m4 because my old computer wasn't having it, since it uses CPU on a different scale than my old version of Logic. But sometimes I run into stuff with some very sketchy timing. This one track I have been messing around with is odd enough where the official Beatport upload of it says it is 125bpm, where the BPM Meter in Logic shows it is dead at 150. It's like one of those "core subgenre tracks", which were always at 150bpm, and I've never seen Logics bpm meter be wrong. It's just arranged to seem that way, but it's the same base problem just for a different purpose. For those kinds of bpm changes, on a full track, which is all I tested to so far in Ripx, it is basically transparent to the ear anyway. It's also a pretty bizarre looking DAW. I asked GPT to explain it as to shorten my post typing it out->

Spoiler

Most time-stretching tools are still fundamentally trying to manipulate a waveform. Even when they're very sophisticated, they're asking, "How do I make this audio longer without sounding terrible?"

RipX is closer to asking, "What notes and sounds are actually in here?" and then operating on those objects. That's why it can do things that seem almost impossible compared to conventional editors.

For transcription, that's a huge advantage. If you're trying to figure out:

• a bassline buried in a mix,
• stacked synth chords,
• a guitar part under vocals,
• complex piano voicings,

being able to isolate, mute, slow, or visually inspect notes is often more valuable than having the absolute best time-stretch algorithm.

It's also why RipX briefly had a reputation for feeling like science fiction before AI stem separation exploded. There was a period where people would load something into RipX and think, "Wait... it can actually see the notes?" Meanwhile most DAWs were still stretching blobs of audio.

That doesn't mean it's magic. You've already run into the reality that dense material can still create "MIDI spam" and note fragmentation. But for analysis and transcription, it's one of the few tools where the workflow itself changes, not just the quality of the stretch.

If the guy's goal is transcription rather than production-quality audio, I'd put RipX near the top of the list because it gives him multiple angles of attack: note display, stem separation, pitch analysis, and tempo manipulation all in one environment. That's often more useful than getting a 2% cleaner time stretch out of a traditional DAW.

Here's a video of a guy doing transcription with it, there are a lot of them; you can at least see the interface and someone's process using it. You can definitely see how well it does polyphonic note detections with the piano and guitar parts:

 

Last edited: Jun 6, 2026 at 5:41 PM