Discontinuous, and spectral selections

[Paul L]

Musing today about features I have heard of in other sound editors...

What if I could drag a rectangle over part of spectrogram view to select a region, but also communicate information about the top and bottom frequencies to an effect, built in or Nyquist? Not likely convenient for multiple track selections.

What if I could select multiple intervals and apply an effect repeatedly to each with a single command? Such as, by first defining labels, then selecting the label track or dragging over part of a label track or tracks.

Have these ideas arisen before?

[Gale Andrews]

What if I could drag a rectangle over part of spectrogram view to select a region, but also communicate information about the top and bottom frequencies to an effect, built in or Nyquist?

Do you mean a full-height selection? People have asked to select arbitrary frequencies such as 1000 Hz to 2000 Hz then apply an effect only to those frequencies.

What if I could select multiple intervals and apply an effect repeatedly to each with a single command? Such as, by first defining labels, then selecting the label track or dragging over part of a label track or tracks.

Audacity already implements the "such as": Edit Menu: Labeled Audio . What we would like to do is "real" multiple selections without labels.


Gale

[Paul L]

I was unaware of these commands. They don't quite do what I imagined but make a workaround possible:

Split-cut at the labels.
Copy into new track.
Select all in the track, control-drag with the star tool to realign.
Apply effect to the new track.
Mix and render with the original track.

That is a start, but how can this be simplified to "apply effect to labelled audio"?

It is also not quite right, because when the effect is applied to the new track, it cannot distinguish the silences between clips from the clips. An effect that does not map silence to silence will create sound in between.

[Paul L]

People have asked to select arbitrary frequencies such as 1000 Hz to 2000 Hz then apply an effect only to those frequencies.

Gale

Can you point me to a detailed discussion of that? I am not exactly sure what is meant by the brief description.

[Gale Andrews]

People have asked to select arbitrary frequencies such as 1000 Hz to 2000 Hz then apply an effect only to those frequencies.

Gale

Can you point me to a detailed discussion of that? I am not exactly sure what is meant by the brief description.

It means if you have a spectrogram showing for example 8000 Hz at the top to 0 Hz at the bottom, you could take your mouse, select a rectangle with a time selection on the x-axis and 1000 to 2000 Hz on the y-axis, and manipulate only those selected frequencies in that time range, for example, delete them.


Gale

[Gale Andrews]

I was unaware of these commands. They don't quite do what I imagined but make a workaround possible:

Split-cut at the labels.
Copy into new track.
Select all in the track, control-drag with the star tool to realign.
Apply effect to the new track.
Mix and render with the original track.

That is a start, but how can this be simplified to "apply effect to labelled audio"?

At the moment you can't apply effects directly to labeled audio. We're discussing a new command to "Select labeled audio" but that is actually an extra step if you want to apply the current Labeled Audio commands.

If you place the cursor in the new track at the start of the first label, you don't need to realign the paste.

It is also not quite right, because when the effect is applied to the new track, it cannot distinguish the silences between clips from the clips. An effect that does not map silence to silence will create sound in between.

If you apply Nyquist effects to the selection including the clips, that will create silence between the clips. Is that what you mean? It is not a problem with built-in effects.


Gale

[Paul L]

It means if you have a spectrogram showing for example 8000 Hz at the top to 0 Hz at the bottom, you could take your mouse, select a rectangle with a time selection on the x-axis and 1000 to 2000 Hz on the y-axis, and manipulate only those selected frequencies in that time range, for example, delete them.


Gale

If you mean "manipulate" very generally, I suppose this isn't so simple as it sounds to implement. I was thinking more modestly of a convenient "notch this!" action. You select a rectangle, and some Nyquist effect might calculate the right central frequency and q.

I have written an effect like that for my own use that figures out the peak frequency in a selection, but sometimes, say, there is one ringing overtone that is not the peak frequency and I want to cut it out. It can be figured out with the help of Analyze, Spectrum but I was thinking of simplifying it.

[Paul L]

If you apply Nyquist effects to the selection including the clips, that will create silence between the clips. Is that what you mean? It is not a problem with built-in effects.


Gale

Suppose I want to generate brown noise (a good approximation to room tone) just in those places. No, instead I get it throughout.

[steve]

It is also not quite right, because when the effect is applied to the new track, it cannot distinguish the silences between clips from the clips.

This is a limitation of the current implementation of Nyquist in Audacity. There has been some preliminary work done by one of the Audacity developers for "version 4" Nyquist plug-ins which don't have this limitation.

Suppose I want to generate brown noise (a good approximation to room tone) just in those places. No, instead I get it throughout.

"Generate" commands are handled differently from "Effects" ("Process" commands).
"Effects" are intended to take the existing audio clip(s) and alter it in some way.
"Generate" commands are not intended to act on the current audio, but on the selection. Thus generating into multiple audio clips (in the same track) is not currently supported, but should become a supported feature when Audacity supports discontinuous selections. However there are a couple of possible workarounds:

1) Label the parts that you don't wish to generate onto.
2) Select the entire track.
3) Create a new audio track.
4) Generate the required audio into the new track.
5) Select both the label track and the new track.
6) Split Delete.

If you are only able to label the parts that you do wish to generate into:
1) Label the parts that you wish to generate onto.
2) Select the entire track.
3) Create a new audio track.
4) Generate the required audio into the new track.
5) Duplicate the new track.
6) Select the duplicate track and the label track.
7) Split delete.
8) Invert the duplicate track.
9) Select the two tracks that contain the generated sound (one of which is continuous sound, the other has multiple audio clips and is inverted).
10 Tracks > Mix and Render.

or Nyquist plug-ins, this could be a lot simpler if Nyquist plug-ins were able to read label tracks. This is another feature which will hopefully be implemented in version 4 plug-ins.

[steve]

People have asked to select arbitrary frequencies such as 1000 Hz to 2000 Hz then apply an effect only to those frequencies.

Gale

Can you point me to a detailed discussion of that? I am not exactly sure what is meant by the brief description.

Not a detailed discussion, but this is the most recent request that I recall seeing for this type of feature: http://forum.audacityteam.org/viewtopic ... 15#p245215