Feature Request: Soft-Border Cut

[Robert2413]

I would like to see an option to allow crossfades across cut (CRTL X) boundaries, which would make life very easy when doing things like seamlessly shortening the gaps between performances in live recordings. Instead of a hard cut, the new option would crossfade the material before and after the cut.

The duration and shape of the crossfades would be adjustable; constant voltage and constant power are the two logical choices.

There are several ways to work out the details. I think any one of them would be satisfactory. One would be to let the borders of the cut define where the crossfades end. Another would be to let the borders define where the crossfades begin. The former would guarantee that any material inside the cut area was not heard, so I think it's probably preferable.

Another application of a soft-border cut is as an alternative to the Interpolate command, particularly when the amount of material removed is more than the Interpolate command can accommodate without audible artifacts. The soft-border crossfade would remove a little bit of the program time, but in many cases this is acceptable.

In a multitrack music production program that I use to create music (as opposed to restoration, which I do in DC7), I use crossfades all the time to create composite vocal tracks from multiple takes. In many cases, the crossfade makes the difference between a glitchy edit and one that sounds totally smooth. Indeed, the old analog tape splicing blocks allow cutting tape at a 45 degree angle for a very good reason -- it created a crossfade at the splice.

[Craig Maier]

DC7 has a crossfade feature; details can be found in the users manual. Is this not sufficient?

[Robert2413]

DC7 has a crossfade feature; details can be found in the users manual. Is this not sufficient?

It sort of awkward if I understand it correctly. Correct me if I'm wrong, but shortening a gap between performances by 5 seconds would require capturing the entire remainder of the concert (even if it's 60 minutes) on the Clipboard and then Crossfade-Pasting it 5 seconds earlier in the file. Because of the large amount of data put on the clipboard and the need for DC7 to write an Undo file for the operation, there can be substantial delays before the operation is completed. Assuming that there are no markers in the rest of the file yet, probably the simplest way to capture the rest of the file is to drop a temporary marker at the beginning of the section to be crossfaded to, double-click after the marker, and invoke CTRL C. The marker would then need to be deleted (although crossfading over it might delete it; I haven't tried it).

My proposal allows you to continue working on the zoomed-in part of the file between performances (there's no need to zoom out to make sure you have captured the rest of the file) and, if programmed efficiently, should be very quick for small cuts (as is already the case for hard cuts) because only a small Undo file needs to be written. So my proposal is really something to speed up the workflow, not something that can't already be done.

[Dan McDonald]

Hi Robert -
I run into this sometimes too, where you want to trim a little bit out but you don't want it to be noticeable. Crossfading crowd noise or whatever makes it virtually unnoticeable.

If you're using classic mode, what happens is that small cuts tend to take a while anyway because the software has to rewrite the whole file. I typically chop each file (like 90 minutes or so) into songs or sections, edit those individually and then do any crossfades if they're necessary, because you'll want to end up with one long file that you chop so that you can quantize for CD.

What I think you're saying is that you'd like to go into the big file and highlight sections that you'd like to trim out, with the option of crossfading the two pieces rather than simply butting them together, right?

I like that option also. I imagine that would have to be a different command because the cut option is nice and simple, so you wouldn't want to have to have a choice there, but if there was a "cut and crossfade" command, that would be great.

It's sort of like the interpolate command but on a bigger time scale.

Dan

Dan

[Craig Maier]

I guess it would be "Crossfade Cut".

What I do in these situations is just cut and then interpolate the seam and that works quite well for me - - - it is usually not audible so long as the cut is less than a few hundred milliseconds long.

[Dan McDonald]

I hadn't thought of just interpolating. That should work pretty well unless someone is in the middle of a word or something.

Dan

[Robert2413]

I guess it would be "Crossfade Cut".

What I do in these situations is just cut and then interpolate the seam and that works quite well for me - - - it is usually not audible so long as the cut is less than a few hundred milliseconds long.

That is a good strategy for short cuts. But as an example of what I would want to do, assume that there is 20 seconds of audience rustle between performances, but the audience is quieting down throughout the 20 seconds. I want to cut 15 seconds out of the middle but avoid the jump between a more active and less active audience by making a slow crossfade between the two.

(As an aside, I think that the current interpolation algorithm could be improved by applying a varying gain function to it that ensures that the power in the interpolation smoothly varies between the power at the interpolation's endpoints. This would prevent the "tape drop-out" sound with large-duration interpolations.)

[Craig Maier]

Quoting Robert:

"I think that the current interpolation algorithm could be improved by applying a varying gain function to it that ensures that the power in the interpolation smoothly varies between the power at the interpolation's endpoints."

------------------------------------------------------------------

How do you know that is not part of what is done presently? We do not divulge our math routines, for obvious reasons. I think that you will find objectively that the Diamond Cut interpolator beats the rest as it is right now with whatever we use. I will point out that part of the routine uses up to 100th order for the curve - fit portion (which is something that is revealed in the product users manual).

[Craig Maier]

Audyossey,

I think that it is fine that you are fond of competitors products and their feature set. That is certainly your perogative. What is not your perogative is to come onto this venue and proclaim all of the virtues of your favorite competitors editor. It has the de-facto effect of bashing the Diamond Cut product line. In fact, I think that it takes quite a bit chuptzah (nerve) to do what you had attempted to do here on this thread. If you want to post those kinds of things, please do so on other websites or audio forums.

We will not allow this forum to degenerate into a Wild West shootout at the OK coral.

Thanks in advance for your cooperation,

The Management

[Robert2413]

Quoting Robert:

"I think that the current interpolation algorithm could be improved by applying a varying gain function to it that ensures that the power in the interpolation smoothly varies between the power at the interpolation's endpoints."

------------------------------------------------------------------

How do you know that is not part of what is done presently? We do not divulge our math routines, for obvious reasons. I think that you will find objectively that the Diamond Cut interpolator beats the rest as it is right now with whatever we use. I will point out that part of the routine uses up to 100th order for the curve - fit portion (which is something that is revealed in the product users manual).

I don't know what the Interpolate algorithm is. I meant my post to be a positive suggestion on how to make it better based on my using it a lot, not only for vinyl restoration, but also to restore live concert tapes. (My current project is restoring about 10 hours of tapes of chorus and band concerts at my high school. These tapes date to the early '60s.)

My observation is based on noticing that the function, when used over long time spans toward the outer limits of what the program permits, tends to cause a drop in loudness in the middle of the interpolation. This is clearly visible when you look at the waveform (the envelope in the interpolated region exhibits causes a sway-back characteristic) and sounds like a tape drop-out. When seeing a pattern like that, the first thing that comes to mind is applying a time-varying gain to the interpolated area that is complementary to the "sway-back"

That being said, the DC7 interpolation algorithm is a good improvement over the algorithm in DC6. I find the algorithm works very well for most impulse noise, particularly in vinyl restoration. But when one is editing live concert material where there may be longer noises, there is often a need for interpolation that works over several hundred milliseconds because of reverberation within the venue.

If you prefer to continue this discussion via private email, that's fine with me.

[Craig Maier]

Quoting Robert:

"This is clearly visible when you look at the waveform (the envelope in the interpolated region exhibits causes a sway-back characteristic) and sounds like a tape drop-out. When seeing a pattern like that, the first thing that comes to mind is applying a time-varying gain to the interpolated area that is complementary to the "sway-back""

-------------------------

The cause of the decrease in accuracy for extremely long interpolations is not due to our fade-in/out process. Your suggestion will not improve things because we already do that. The problem is that the routine is limited to 100th order curve-fits and at some point, it has a tough time anticipating what should have been in the highted area of a file. That is a natural by product of these types of statistical calculations. There will always be a time limit beyond which one can not accurately interpolate a file; the longer the highlighted portion of a file, the more inaccurate the interpolation. As an example, imagine interpolsting an entire symphony give one second of it at the beginning and another at its end? I use this to make my point.