Jump to content

Cakewalk 2021.09 Early Access Available


Larry Shelby

Recommended Posts

Interesting looking update, I usually don't do early access, but I might this time. If I read this correctly, exporting individual arranger sections as audio would save me a huge amount of time and effort when exporting cut-downs of mixes that I need to create. Stem export also looks promising, too, as I also need to export stems frequently in my work as a composer. Gonna give this a spin, I thin'

Link to comment
Share on other sites

7 hours ago, Piotr said:

At last! :) They noticed transient detection is kind a ... in CbB... I hope they will do their best. It would be so welcome to have it done perfectly...

 

6 hours ago, chris.r said:

I've never noticed any transient detection issues. Curious to see if you guys find the improvement useful.

 

I saw a difference with the new transient detection, it's slightly better. Though with a very simple tambourine loop the result has been very poor IMO, some transients are quite okay and others are very late within the hits (as with the old standard detection algorithm). With a bass clip the results were more pleasant. It is also interesting that the first of the 3 algorithms (Magnitude, old) showed the best result for the tambourine clip!? ?

Sometimes I wonder what is the correct position for the transient detection, especially when I compare the results of different DAWs and algorithms? ? Also in the internet and in videos you find different answers and descriptions for transients. So I am a bit confused! Nevertheless I am sure that many detections are not so accurate when I listen to the quantized results!

But in general the weak detection of transients in AudioSnap was not the worst problem for me. I did more fight with the results of the render that seemed a bit moody to me. Often it did not follow the transient settings (there were significant moves), sometimes even huge jumps. When I had moved 1 or 2 transients just a little bit (< 5 samples), then suddenly the result was better! But such try and error takes too much time! Another thing that happened to me several times is that the transient part of a note/hit just disappeared with the render or inverse it was repeated (occurred twice in the same note start). Still I would like to use the CbB AudioSnap, because I like that the quantization supports swing and strength (percentage) ?. Most other DAWs don't!

If I compare AudioSnap to Reaper's stretch markers, then I still prefer to work with Reaper! On the one hand the detection is more reliable, second the detection results are just proposals and you can just go thru and confirm the ones you want (click). This goes really fast! Additionally there are very cool possiblities to configure the stretching (not only the algorithm) and there exists also a tape-speed-like stretching that changes the pitch naturally!

 

Link to comment
Share on other sites

20 minutes ago, marled said:

I saw a difference with the new transient detection, it's slightly better. Though with a very simple tambourine loop the result has been very poor IMO...

I've never experienced these issues myself but that's probably because I've never pushed it to it's limits. Are there more apps having better detection/render than Cakewalk? I saw discussion in the EA thread about it and Noel, or Mark, asking for samples. Maybe it would be good idea if you give them your sample with the description as they are working on it now?

Link to comment
Share on other sites

8 hours ago, chris.r said:

I've never noticed any transient detection issues. Curious to see if you guys find the improvement useful.

Did you try do do some work with audio of instrument like guitar for instance? CbB is finding transients in insane places and missing where they are.  Threshold is not enough to control it and manual work to move/add/delete is painful.

I would say none progress with it from ancient times so I am happy to hear anything is going to be better about it :)

The goal should be such good detection to be used in cases like getting audio track to quantize so after use transient detect function and quantize it is done ;)
Not that you need to sit hours to manually adjust.

  • Like 1
Link to comment
Share on other sites

8 hours ago, chris.r said:

Maybe it would be good idea if you give them your sample with the description as they are working on it now?

I am not allowed to exchange this sample, it is a paid one! And anyway, I think I am on their black list! ?

 

7 hours ago, Piotr said:

The goal should be such good detection to be used in cases like getting audio track to quantize so after use transient detect function and quantize it is done ;)
Not that you need to sit hours to manually adjust.

 

7 hours ago, chris.r said:

Yeah I've been always doing the tedious stuff manually and never gave it a second thought ?

Looking at the current state of the art I am not confident that there will be a 100% automatic transient detection! And in some cases (vocals, electric guitar, ...) there is not just one way to define the transients, i.e. there is more than one correct possibility, but the results sound different.

So I believe Reaper's half-automatic detection is the best way to have satisfactory transient results! I am very pleased with it and quantization alone has been worth to invest in Reaper (as said somewhere else I still like to use CbB for other stuff)!

  • Like 1
Link to comment
Share on other sites

2 hours ago, marled said:

Looking at the current state of the art I am not confident that there will be a 100% automatic transient detection! And in some cases (vocals, electric guitar, ...) there is not just one way to define the transients, i.e. there is more than one correct possibility, but the results sound different.

Yup. I am aware of it but anyway reaching consistent and logical placement  should be possible. Then we would need only some adjustments instead of massive amount :)
Especially frustrating is observing when CbB is placing transients marker in places where even blind and drunk man never would place it :( Or skipping some so obvious then almost screaming for it.

Anyway they could also try to add  machine learning in development to improve detector without over-complicating algorithms.  I believe it is a great application for it.

2 hours ago, marled said:

So I believe Reaper's half-automatic detection is the best way to have satisfactory transient results! I am very pleased with it and quantization alone has been worth to invest in Reaper (as said somewhere else I still like to use CbB for other stuff)!

Thanks for the recommendation. I have to take a look at the next time I will struggle with transients. Although still it would be great not to jump over DAWs with different task for the same project :)

  • Like 1
Link to comment
Share on other sites

14 hours ago, marled said:

Though with a very simple tambourine loop the result has been very poor IMO, some transients are quite okay and others are very late within the hits (as with the old standard detection algorithm).

Edit: I did some further testing and managed to get a satisfactory result with the transient detection (setting the resolution to 1/8 instead of all).

 

14 hours ago, marled said:

But in general the weak detection of transients in AudioSnap was not the worst problem for me. I did more fight with the results of the render that seemed a bit moody to me. Often it did not follow the transient settings (there were significant moves), sometimes even huge jumps. When I had moved 1 or 2 transients just a little bit (< 5 samples), then suddenly the result was better! But such try and error takes too much time! Another thing that happened to me several times is that the transient part of a note/hit just disappeared with the render or inverse it was repeated (occurred twice in the same note start). Still I would like to use the CbB AudioSnap, because I like that the quantization supports swing and strength (percentage) ?. Most other DAWs don't!

Edit: Those rendering problems seem to appear only with the "Radius ..." algorithms! Though I like them, because they are usually much more accurate (the zoomed in wave form looks very similar). But there are sometimes displacements greater than 10ms (I mean displacement from the stretched audio to the rendered audio)!

Using the "Elastique Pro" algorithm the displacement of the start/end of the notes/hits is only minor and absolutely acceptable. But this algorithm shows often a drastically changed wave form (degradation) even if there is not going on a lot of stretching! (zoomed in to about -18 dB).

  • Thanks 1
Link to comment
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...