binaural audio finally free (as in freedom)

[RiverStreamSounds]

I do not think that is a solution.
The best average size of shoe will still not fit most people. And ears are more different than feet.
I think Virtuoso is doing it right. Giving the possibility to choose an HRTF from a collection in the app, to search a best personal fit among the available databases or even to have somehow (Genelec, labs, ...) get a personal .sofa created.

I don't disagree that in principle, you ought to be able to select your own HRTF. I will be adding that to this project. But since it's aimed at consumers (not professionals), it nonetheless has to ship with a stock one. It is precisely this one I'm investigating. Do you in fact happen to have access to APL Virtuoso by any chance?

 

[olieb]

Actually I do. But I am on Mac.

But since it's aimed at consumers (not professionals), it nonetheless has to ship with a stock one. It is precisely this one I'm investigating.

I understand. But to me it seems so much better to provide a choice of HRTFs.
When I first tested Virtuoso I was a bit underwhelmed with using KU100 HRTF until I checked the built-in ones. And it was the last (E) that made the difference.
However, still looking for a genuine personal one.

 

[RiverStreamSounds]

Actually I do. But I am on Mac.

I understand. But to me it seems so much better to provide a choice of HRTFs.
When I first tested Virtuoso I was a bit underwhelmed with using KU100 HRTF until I checked the built-in ones. And it was the last (E) that made the difference.
However, still looking for a genuine personal one.

I actually need someone who has access to APL Virtuoso on their system to contribute to this project. If you have some free time I would really appreciate it. Perhaps, I could send you a private message and let you know the type of data I'm trying to collect. I'm going to work on a workflow for Mac systems in order to collect this data so I do not believe it would be a problem. I originally mentioned Windows specifically in the OP because I believed that to be the best solution however that is not necessarily the case based on my further investigation into this.

 

[RiverStreamSounds]

I would like to add as well, that I am also working on a separate project, namely, the extraction of 2D and 3D ear anthropometric measurements from multi-view monocular images. Or in plain English, taking a SOFA file that is otherwise generic, alongside pictures of both of your ears from multiple angles taken using a smartphone in a well-lit environment, and computing a different SOFA file as output, which contains the precise differences between the generic file and the shape of your ears.

For compatibility with consumer use cases, because the pictures of your ears are two-dimensional rather than three-dimensional (i.e there is no real 3D scanning hardware involved in this process), the amount of information that would typically become possible to be extracted with somewhat of a reliable precision is limited. Nonetheless, it may well be the case that these limitations prove to be better than utilizing a stock SOFA file on its own.

However, this is a highly complex endeavor, therefore, it will not be my primary focus for now, until I'm done with the primary project first, and secondly, because I most likely won't be able to do all of this on my own, so I would be looking for DSP engineers who are willing to donate some of their time in order to assist with this task. If not, your contribution to the primary project in the OP is also welcome.

 

[olieb]

I actually need someone who has access to APL Virtuoso on their system to contribute to this project. If you have some free time I would really appreciate it. Perhaps, I could send you a private message and let you know the type of data I'm trying to collect.

I might be able to do something, so go ahead. But why do you not do it yourself?
Virtuoso plugin is only a third (67$) of what it used to be. Best audio investment I ever made - perhaps apart from Salnotes Zero2.
And I am no engineer, so python or command line are not my world.

Or in plain English, taking a SOFA file that is otherwise generic, alongside pictures of both of your ears from multiple angles taken using a smartphone in a well-lit environment, and computing a different SOFA file as output, which contains the precise differences between the generic file and the shape of your ears.

Sounds very interesting. Though I do not understand why a generic file (with "generic ears") is involved.
Why not create the actual individual HRTF? There are already some tools (Mesh2HRTF and so on) available. It is just that the whole process (i.e. cleaning up the mesh) is quite tedious and more for experts.

 

[RiverStreamSounds]

I might be able to do something, so go ahead. But why do you not do it yourself?
Virtuoso plugin is only a third (67$) of what it used to be. Best audio investment I ever made - perhaps apart from Salnotes Zero2.
And I am no engineer, so python or command line are not my world.


Sounds very interesting. Though I do not understand why a generic file (with "generic ears") is involved.
Why not create the actual individual HRTF? There are already some tools (Mesh2HRTF and so on) available. It is just that the whole process (i.e. cleaning up the mesh) is quite tedious and more for experts.

1) I'm planning to extract the data of a lot of tools, so it would not be economical to purchase all of them. However, if somebody already has them and is willing to contribute that will be wonderful. I am utilizing this data for research purposes only, and it's very important to note that any further research gained from this understanding will be composed solely of previously openly available work.

2) The primary reason comes down to the sub-millimeter precision bottleneck. Multi-view monocular images (like photos taken from a smartphone) are excellent for estimating general shapes but are inherently noisy. They struggle to perfectly capture the occluded, complex cavities of the ear. Generating an entire HRTF using those pictures and only those pictures alone would have a lot of reliability problems. Therefore, the eventual plan is to modify an existing generic HRTF with semi-personalized extracted features.

Bear in mind that I would intentionally be quite conservative with this process, so the amount of modifications that would then become possible is naturally going to be limited. However, due to the way our ears work, it's very possible that this small modification could well be enough make a larger difference for somebody out there.

And as I have already mentioned before, this is an order of magnitude more difficult than the existing project I'm working on in the OP, so it will only come later, if it arrives at all. I'm doing all of this alone, as you can see, which, for the second project, will simply not be a viable approach.