High quality audio morphing effect plugin, why it doesn't exist yet?

[monas]

yep, exactly this can be done in metasynth as well, all in the same application. you can analyse an audio file and turn it into an image which is resynthesized in an additive way. you can do this with a second audiofile and then just blend the images together the way you like, using all the copy/paste functionality and painting functionality that metasynth offers.

Are you sure Metasynth supports this way of morphing?
You are speaking about "blend", that's different to "morph" and "warp".

You had to select one area (freehand selection) and another target area (2nd picture, different place, size and shape), than you had to define a time frame to morph from the first shape to the second different shape and render this as your output.
Have you tried WinMorph? There are also some basic tuts included that show the process.
btw, I can't test Metasynth, I'm on Windows.

yes, there's all kinds of ways of how to combine pictures in metasynth when you copy them on top of each other (multiply, merge, add, subtract,...). i was just calling it "blend", you can call it "morph" if you wish
you'd have to fade one sound file in, and the other one out (you can do that in the graphics editor if you wish). then you copy the images on top of each other with the fade parts overlapping.

plus you can lay hand on the image yourself by painting into it to edit things and make the transition better. and you have the two fundamentally different options of resynthesizing the file in an additive or substractive way. Metasynth is a really deep tool. i have to say i never really used it for morphing sounds, but it is surely capable of it.

[highkoo]

You guys gave me another piece of a half an idea...

Go to the websites for various Adobe PS/AE plugins,
Dump their sample before/after pics into Metasynth and see what the plugins can 'sound' like.

[valhallasound]

You guys gave me another piece of a half an idea...

Go to the websites for various Adobe PS/AE plugins,
Dump their sample before/after pics into Metasynth and see what the plugins can 'sound' like.

Working in Metasynth, you won't want to use any video morphing plugins. Metasynth represents a sound of X duration as a single image, so you would want to use a filter that could somehow "interpolate" points between the left and right. I'm not sure what that would look like.

I wonder if there are more of these morphing type tools on OSX, that derive from the work done in the 1990s on Macs. Back when I was young(-ish) and poorer, I used Csound on a Pentium II running Windows 98 to create my sounds. I always had serious envy of those folks that had Macs, as they could run Soundhack, Supercollider, Metasynth, Max/MSP, and so on. All of these tools are still available, and I think that Soundhack and Metasynth are OSX only.

Sean Costello

[VariKusBrainZ]

I think Fscape might do morphing, been so long since I played with it.
Has lots of interesting modules. Inspired / similar to Soundhack

http://www.sciss.de/fscape/

[deastman]

You guys gave me another piece of a half an idea...

Go to the websites for various Adobe PS/AE plugins,
Dump their sample before/after pics into Metasynth and see what the plugins can 'sound' like.

Working in Metasynth, you won't want to use any video morphing plugins. Metasynth represents a sound of X duration as a single image, so you would want to use a filter that could somehow "interpolate" points between the left and right. I'm not sure what that would look like.

I wonder if there are more of these morphing type tools on OSX, that derive from the work done in the 1990s on Macs. Back when I was young(-ish) and poorer, I used Csound on a Pentium II running Windows 98 to create my sounds. I always had serious envy of those folks that had Macs, as they could run Soundhack, Supercollider, Metasynth, Max/MSP, and so on. All of these tools are still available, and I think that Soundhack and Metasynth are OSX only.

Sean Costello

I agree. Metasynth really doesn't do morphing, just crossfading between two images. If it did visual morphing, it would have to warp the images over time so that similar features in the two images gradually aligned over time as the "blend" happened. To make this happen, you would have to create a video morph in another program, and Metasynth would have to be capable of importing a video sequence and resynthesizing a different video frame at each time slice of the audio. Not to mention that video typically only refreshes at 30 frames per second, as opposed to 44.1khz or greater for audio.

Its a misconception to think that existing visual image processing tools can be reapplied to the audio domain and magically produce a morphing tool which exceeds the results already attainable by existing audio tools.

[deastman]

Harry_HH: The solution to your lack of a good morphing tool is obvious. You should stop complaining and just go program one yourself. Or buy a Kyma system.

You're getting the same response as when a 5-year old demands a rocket-powered unicorn.

OMG! I want one of these sooo bad!

[monas]

You guys gave me another piece of a half an idea...

Go to the websites for various Adobe PS/AE plugins,
Dump their sample before/after pics into Metasynth and see what the plugins can 'sound' like.

Working in Metasynth, you won't want to use any video morphing plugins. Metasynth represents a sound of X duration as a single image, so you would want to use a filter that could somehow "interpolate" points between the left and right. I'm not sure what that would look like.

I wonder if there are more of these morphing type tools on OSX, that derive from the work done in the 1990s on Macs. Back when I was young(-ish) and poorer, I used Csound on a Pentium II running Windows 98 to create my sounds. I always had serious envy of those folks that had Macs, as they could run Soundhack, Supercollider, Metasynth, Max/MSP, and so on. All of these tools are still available, and I think that Soundhack and Metasynth are OSX only.

Sean Costello

I agree. Metasynth really doesn't do morphing, just crossfading between two images. If it did visual morphing, it would have to warp the images over time so that similar features in the two images gradually aligned over time as the "blend" happened. To make this happen, you would have to create a video morph in another program, and Metasynth would have to be capable of importing a video sequence and resynthesizing a different video frame at each time slice of the audio. Not to mention that video typically only refreshes at 30 frames per second, as opposed to 44.1khz or greater for audio.

Its a misconception to think that existing visual image processing tools can be reapplied to the audio domain and magically produce a morphing tool which exceeds the results already attainable by existing audio tools.

true, metasynth doesn't have a dedicated algorithm for image morphing (interpolation of images), but it does A LOT more than just crossfading. it would be worth a feature request though. it shouldn't be hard to implement.

but if you edit the images by hand and do some trickery it should be possible to get there still. also there's the spectral room which offers a third possibility for these kind of manipulations. it would take some time and some skill to make a morph sound good in metasynth.

[tweiss2000]

Its a misconception to think that existing visual image processing tools can be reapplied to the audio domain and magically produce a morphing tool which exceeds the results already attainable by existing audio tools.

Why should visual processing be misconception?
We are all used to optical representation of sounds, wav view, spectral view...
And why not use the same direction for morphing sounds?
E.g. Izotope iris uses selection of optical represented parts of audio to derive new sounds.

Iris could also be a good basis for morph.
It would need to add a second wave form as target wave, a function to assign the source to target parts, and finally a rendering over time.

A further direction could be a photoshop plugin to support audio and morph.
Or is such a plugin already available?

But all in all I would love a vsti, integrated in my daw as I'm used too. Complex envelopes for morphing, layering of multiple waves, integrated editing, blending, filtering of layers ...just dreaming about that

Beside that dream I'm now comming back to Alchemy, as I don't have the impression, that there is any developer, who jumps into developing that dream.

[whyterabbyt]

Why should visual processing be misconception?
We are all used to optical representation of sounds, wav view, spectral view...
And why not use the same direction for morphing sounds?

because they're all visual abstractions of the relevant audio rather than being the relevant audio. manipulating the visual abstraction in a purely visual manner doesnt have a 1:1, or even predictable, correspondence with what that does to the audio, because its an abstraction.

For example a visual 50/50 'fade' between a waveform view of a sine and a square is not going to have the harmonic content of a 50/50 crossfade between a sine and a square.

E.g. Izotope iris uses selection of optical represented parts of audio to derive new sounds.

But there's nothing visually in that selection which allows you to predict what you're going to hear. You actually have to hear it to know what you actually selected... because its an abstraction.

Thats not to say that morphing of the visual abstraction might produce interesting audio results, but they're far less likely to be representative of audio morphs than you'd want.

Iris could also be a good basis for morph.
It would need to add a second wave form as target wave, a function to assign the source to target parts, and finally a rendering over time.

I agree that the type selection tools in Iris could serve morphing well. Im not entirely sure if Iris would be a good starting point, though.

A further direction could be a photoshop plugin to support audio and morph.
Or is such a plugin already available?

Not that I can think of. I dont believe the photoshop SDK handles audio, since photoshop doesnt.

If I was looking for a solution for this myself, I probably wouldn't be looking at a realtime audio plugin, I'd look to building something standalone, and in terms of building on what's already out there, I'd start with MAX/MSP front end to the various CSound opcodes.

aside : Tux, you out there? How did you get on with the loris opcodes I built for you?

[monas]

because they're all visual abstractions of the relevant audio rather than being the relevant audio.

??
there's no difference. of course digital audio can be perfectly represented in an image. it can be a spectrogram, which can be exactly played back just like a time domain based wave form display can. that's what for example metasynth is all about.

especially for morphing it is pretty much unavoidable to analyze to frequency domain (spectral representation) and resynthesize that "image".

[whyterabbyt]

there's no difference. of course digital audio can be perfectly represented in an image. it can be a spectrogram, which can be exactly played back just like a time domain based wave form display can. that's what for example metasynth is all about.

no, the transliteration or abstraction of audio data into a visual form doesnt mean that the visual form is the same as the audio data.

and just because something gets played back that doesnt mean its the visual data that gets played back.

crudish example : in any audio editor, there's a finite resolution at which waveforms get displayed visually. have you ever seen one which allows you to zoom in to a per-pixel level of detail which matches the 24-bit resolution of common audio files, with the display 2^24 pixels high?
if not, that visual representation is not same as the underlying audio data.

second example : does a spectrogram display the phase of partials?

anyway, thats all beside the point. whether i'm right or wrong about that, its not my underlying point. my point is that manipulating the visual abstraction in a particular manner (ie 'visual morph') does not result in a 1:1 correlation with the effect on the underlying audio data (ie 'audio morph').

[monas]

there's no difference. of course digital audio can be perfectly represented in an image. it can be a spectrogram, which can be exactly played back just like a time domain based wave form display can. that's what for example metasynth is all about.

no, the transliteration or abstraction of audio data into a visual form doesnt mean that the visual form the same as the audio data.

and just because something gets played back that doesnt mean its the visual data that gets played back.

crudish example : in any audio editor, there's a finite resolution at which waveforms get displayed visually. have you ever seen one which allows you to zoom in to a per-pixel level of detail which matches the 24-bit resolution of common audio files, with the display 2^24 pixels high?
if not, that visual representation is not same as the underlying audio data.

second example : does a spectrogram display the phase of partials?

especially for morphing is pretty much unavoidable to analyze to frequency domain (spectral representation) and resynthesize that "image".

but i MEANT the VISUAL data getting played back. and of course that's possible. of course there's a loss of certain information after a fourier transform, and it won't be sample accurate to the original, but that's a price you have to pay if you do spectral domain processing.

[monas]

second example : does a spectrogram display the phase of partials?

phases can be completely ignored for resynthesis

[whyterabbyt]

second example : does a spectrogram display the phase of partials?

phases can be completely ignored for resynthesis

'Can' or 'will' ?

And this is still quibbling over things which werent my actual point. Manipulations of the asbtraction of the data dont necessarily correlate in any meaningful manner with the resultant effect on the original data. Are you disputing that, or not?

[monas]

i really don't know what you're trying to say, but if you choose to believe that spectral processing isn't possible, you're free to do so.