Skip to main content
Participant
August 7, 2017
Question

User-guided audio selection from complex sound mixtures

  • August 7, 2017
  • 3 replies
  • 595 views

Hi,

quite some time ago I found an interesting program and it turns out Adobe owns the rights to this software. The patentnr. is US 8954175 B2.

Is there any chance that this will be implemented or maybe it already is and I don't know it?

Kind regards

Phil

    This topic has been closed for replies.

    3 replies

    PhilP88Author
    Participant
    August 9, 2017

    These examples are made just to make this tool look good. The same thing with the siren, you can clearly hear that they added that on top of the original track. I haven't seen a demonstration of this tool besides the videos from Adobe.

    Bob Howes
    Inspiring
    August 9, 2017

    Between the Sound Remover and tricks you can do in spectral view, I'd agree with Ryclark that the bits that actually work are already incorporated in Audition.

    You mention a siren.  Well, a couple of years ago somebody posted the REAL audio from a wedding video that was spoiled by a passing siren--and by careful use of the tool in Spectral display I was able to completely eliminate it with no nasty effects on the audio of the ceremony.  Similarly, I've totally removed vocals in situations where the vocalist was panned centrally (with no reverb) and other instruments were panned elsewhere.  Unfortunately these conditions rarely apply.

    Finally, for removing room reverb, over the years I've seen several demos of software claiming to do this--and in every case, the results sounded worse to me--reverb still there and lots of nasty artefacts.

    As for unqualified opinions, I don't share Steve's MSc but do have a 35 year career in audio for broadcast TV, first actually doing it and later managing the department--including making hardware and software choices.

    PhilP88Author
    Participant
    August 8, 2017

    Have you watched the video? I'm interested in a reply from Adobe itself, if possible, not some guy who doesn't know anything about the subject but likes to get his unqualified opinion out there. The unmixing a cake thing comes from people who have no clue about source separation, just look at ADX Trax, which I have. It works incredibly well but only for monophonic sources.

    ryclark
    Participating Frequently
    August 8, 2017

    What video is that? I don't see a link one or even what this thread is about.

    SteveG_AudioMasters_
    Community Expert
    Community Expert
    August 8, 2017

    Quite. As for the 'unqualified opinion' bit, I'd say that whoever this guy is, he's an expert on making those. As for me, I have an Msc in Acoustics...

    Anyway, the reason that it won't work successfully is that with any significant reverberation in a congested sound field, all the temporal information is smeared, so what (if anything) you extract in one instant by deconstructing the harmonic content will leave a significant part of its content on what immediately follows it - which is why de-reverb is not really any sort of success.

    Now let's consider ADX Trax. It's intended to be a vocal removal tool, and I've only seen demos where they pick a really easy song to deconstruct, where the vocal is spatially removed, at least in the frequency spectrum.

    Now, what this patent claims to be able to do is pretty much what ADX Trax does, and there are two things to consider about that; the first is that just because you make a patent claim, it doesn't give you any ultimate rights to use that technology exclusively, especially if you neglect to mention any prior art. Now, that's not the case here; thirteen other prior publications have been mentioned, and I'm reasonably sure that as a result of considering the content of some of them, that any challenge to this patent would stand up; the knowledge-base behind this technology is effectively in the public domain, which is probably why Adobe haven't made any sort of a fuss about it.

    There are other audio examples of this sort of patent malpractice. My favourite one is this: Open Ambience Project "SHEPPi" Free Spatial Enhancer. It happens because the patent office simply doesn't have time to check all the applications they get rigorously, but they are still quite keen to take your money. So are lawyers, so any legal contest - well it's win-win for all of them (but not you).

    But hey, I listened to the ADX Trax results, and I'm pretty underwhelmed. For the amount of fiddling about involved, I could do that just as well with a combination of Audition and iZotope's RX6. And in stereo.

    SteveG_AudioMasters_
    Community Expert
    Community Expert
    August 7, 2017

    No it isn't, and I don't for one moment think that it will work very well. About as well as reverb removal, in fact - for similar reasons.

    The other, and quite correct, way to look at it is that if it was that good, why would they have been sitting on it for eight years?

    Bob Howes
    Inspiring
    August 8, 2017

    Steve, you forgot your un-mixing a baked cake synonym...

    SteveG_AudioMasters_
    Community Expert
    Community Expert
    August 8, 2017

    I was saving that!