Hello,
I’m still not an SL user. I’m wondering if the program could remove the Bb note (which is out of key, for sure) that we can hear in the following file, please.
If so, how to do it?
Thank you.
Impressive! How did you do this?
Not very well done though.
Unmix song, choose guitar and piano then mute the layer “other”. I had to erase the tail as that was still Bb in the reverb but you can do it better on that layer selecting the tail and using the denoise tool.
Step-by-Step Guide
-
Change the Frequency Scale to MIDI-Logarithmic
Open the Frequency Scale settings and switch to MIDI-Logarithmic view.
This mode displays frequency bands as musical notes, making it easier to identify and edit specific pitches.
-
Increase FFT Size to 24,576 or Higher
Set the FFT (Fast Fourier Transform) size to 24,576 or more for higher frequency resolution.
This provides a more detailed spectral view, allowing you to pinpoint and isolate notes precisely.
-
Zoom In and Select the Target Note (Bb)
- Vertically zoom in to focus on the specific note you want to remove.
- SpectraLayers labels notes using fixed enharmonic symbols (e.g., A# instead of Bb, D# instead of Eb).
- In this case, to remove Bb, look for A# on the frequency scale.
- Use the Rectangular Selection Tool to select the unwanted note region, then press Delete.
-
Review the Result
After deletion, you’ll see that the fundamental (lowest) component of the note has been removed.
-
Remove Remaining Harmonics (If Needed)
- The deleted note’s harmonics may still be present at higher octaves (e.g., A#5 and above).
- To fully clean the note, repeat the process for each harmonic manually, or use the Harmonic Selection Tool to remove all related overtones automatically.
- For guitar recordings, harmonics are often close to octave multiples of the fundamental, but this can vary depending on the string and pickup characteristics.
6 Likes
Thank you very much for your replies. SL seems to be a great tool. I will buy it soon.
A nice post; I particularly appreciate guidance on FFT size.
This is not the only way to edit in SL; this description is a destructive process. If user wants to back out of the editing performed, using destructive processes such as eraser tool and delete function remove the selected and If user wants to undo beyond the SLP per session undos and their limitations using a muted trash layer will allow non-destructive editing.
1 Like
Thanks for this important precision.
1 Like
well, we could go on and on about SLP various selection processes. Making a single rectangular selection and repeating that tool usage over and over is but one way to work. Understanding the selection modes to use them in combinations with various tools to create complex selections is the essence of SLP.
1 Like
Additionally, the window size function is very dependent on the specificity of the goal. A window of 24000+ samples is typically only useful for low frequencies and actually reduces temporal accuracy for mid and high frequencies. If the note duration isnt long, you dont need a huge FFT window. It’s often more useful to use a mid-size window (say, 4096 samples) for the task in this thread, but use the resolution and refinement sliders to increase the accuracy of the graphic. Usually you only need longer FFT windows for working with long tones or low bass notes.
For working with transient or short sounds, short FFT windows are much better. For clicks or pops, you really only need windows from 512 to 2048 samples depending on the frequency range of the transient anomaly.
Using an inappropriate window size can cause too much injury to audio surrounding the anomaly you are working to eliminate. Always best to try multiple window sizes until you become very familiar with choosing the optimal window size for the task.
You can easily try adjusting the 3 sliders to see the resolution and refinement parameters work with the graphic display.
4 Likes
My default is 2048 for editing NR: human voice, animals and insects, vehicle sounds, any fx, all unwanteds as well; extreme lower or higher settings I have never arrived at a use for…I’m happy to learn what others do and would love to see some video explanations for various alternate usages 
It might also be useful to consider the samplingrate when adjusting the FFT-size.
yes. Any parameter which affects accuracy of representation, editing precision and minimization of artifacts. This is one of the nice things about spectral editing in the graphics domain - the drawing usefully displays the results of the transform parameters.
The math is convoluted 
Try searching for terms like fourier transform windowing, short time fourier transform, wavelet transform, measuring time domain vs frequency domain
1 Like
no, I mean video as pertains to use specifically in SL…concepts is one thing (ie learning to read the spectrogram)
I don’t want to seem pedantic by saying RTFM, but if you want to tweak the parameters which affect the transform from audio to graphic data, you really need to understand the basics of the methods used. Robin explains it simply in the SL docs, but he does the user no favours by not being explicitly clear in the GUI regarding which processes are subject to transform parameters. Instead of making this clear in the menus, he provides a list in the docs. Very unintuitive for users who don’t know much about transforms!
Izotope does a much better job at this in the UI for RX and the associated docs. You will probably find videos which examine these aspects if you search for RX guides.
The main thing you need to know is the relationship between time domain and frequency domain.
The window size not only affects the initial transform, it also affects the derived solution because the chosen parameters are used for the fix applied to the anomaly (assuming you apply a spectral process).
For editing short transients such as clicks, a small window is better. For editing longer sounds such as sustained tones, larger windows are usually better. However, this depends on the frequency - bass frequencies typically need larger windows than (say) a 1kHz tone.
Below are some screenshots of the same audio data at different block sizes:
2 Likes
I requested this already sometime earlier.
It would also be nice, if the FFT-Size could be greater than 32768 samples.
I get it, man…I have RTFM over and over and over and over
I think I can read the spectrogram pretty well, but maybe not…it’s not like I rub shoulders with audio scientists daily…or at all…I’m all self-taught
I specifically would enjoy seeing videos on real usage in SLP
In my many, many hours in SLP over the past 14 months I have have very, very little effective usage of changing FFT size beyond my default pref of 2048, if any. That said, I repeat, I don’t edit much music.
I don’t comprehend the term “transforms” as a noun related to FFT size.
I can’t say I agree with this in my usage. It’s kind of like binoculars, either the image is clear or it ain’t
Fast Fourier Transform = FFT
It’s the method which converts the audio data into graphic spectral data, which is the core job of the application
Again, you just show you don’t know what you don’t know.
Using a block size of 2048 samples for low frequencies is not optimal at all. I just showed you why.
yes, the beauty of spectral editing is you can visually refine the settings. You can even do things at inappropriate block sizes. But you won’t derive optimal (ie the best) results unless you choose the right block size for the task.
I’ve only been doing spectral editing since 2004 and I _still_ do regular research on methodology. You? You ask for others to provide you with videos.
I’m out of this thread.
2 Likes
wow, thanks for the lesson. I am a newbie, an old man, but new to spectral editing. I do not know a lot, that is for sure. As I proved my ignorance on obvious things. I will take your advice and review the materials you suggested.
I was in no way meaning to make things personal, I apologize for upsetting you, this was not my intention.
I just said I would like to see videos of SLP as related to FFT settings…I didn’t mean to ask them to be made for me.
I guess this thread has diverted plenty.
peace
2 Likes
ah, ok, so, I had a section that I did poorly early on; main reason I put this audio in SLP was unwanted LF wind noise. So, I need to fix some mistakes with my early teething. Yes, indeed this LF info appears smeared at 2048. At 6144 is much, much easier to see what is going on.
Thanks for correcting me, I really do appreciate it
1 Like