1. Forum
  2. Usenet
  3. REC.AUDIO.PRO

  • AI is coming to audio - see Adobe's demo

    From Jason@21:1/5 to All on Tue Dec 20 13:46:44 2022
    Adobe released a demo (https://podcast.adobe.com/enhance)of an
    AI-based audio enhancement routine for voice. You can try it at the link
    above. I recently recorded a chamber music concert at a local church.
    One of the performers spoke for a minute about a piece they were about
    to play. The recording has quite a bit of reverberation and random
    noises. The result from Adobe is surprisingly good, 'tho my wife
    commented that the process altered the timbre of the narrator's voice,
    but not in a way that someone who didn't know her would notice.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From geoff@21:1/5 to Jason on Wed Dec 21 09:54:25 2022
    On 21/12/2022 7:46 am, Jason wrote:
    Adobe released a demo (https://podcast.adobe.com/enhance)of an
    AI-based audio enhancement routine for voice. You can try it at the link above. I recently recorded a chamber music concert at a local church.
    One of the performers spoke for a minute about a piece they were about
    to play. The recording has quite a bit of reverberation and random
    noises. The result from Adobe is surprisingly good, 'tho my wife
    commented that the process altered the timbre of the narrator's voice,
    but not in a way that someone who didn't know her would notice.

    Certainly good enough (excellent ?) for its main declared purpose.

    Although in many cases the background environment imparts a context and authenticity to, say, an interview or news report.

    geoff

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Orlando Enrique Fiol@21:1/5 to All on Fri Dec 1 17:49:11 2023
    In article <[email protected]>,
    Adobe released a demo (https://podcast.adobe.com/enhance)of an
    AI-based audio enhancement routine for voice. You can try it at the link >above. I recently recorded a chamber music concert at a local church.
    One of the performers spoke for a minute about a piece they were about
    to play. The recording has quite a bit of reverberation and random
    noises. The result from Adobe is surprisingly good, 'tho my wife
    commented that the process altered the timbre of the narrator's voice,
    but not in a way that someone who didn't know her would notice.
    What actually constitutes timbral alteration knowledge?

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Scott Dorsey@21:1/5 to [email protected] on Sun Dec 3 14:15:55 2023
    Orlando Enrique Fiol <[email protected]> wrote:
    In article <[email protected]>,
    Adobe released a demo (https://podcast.adobe.com/enhance)of an
    AI-based audio enhancement routine for voice. You can try it at the link >>above. I recently recorded a chamber music concert at a local church.
    One of the performers spoke for a minute about a piece they were about
    to play. The recording has quite a bit of reverberation and random
    noises. The result from Adobe is surprisingly good, 'tho my wife
    commented that the process altered the timbre of the narrator's voice,
    but not in a way that someone who didn't know her would notice.
    What actually constitutes timbral alteration knowledge?

    For voice that's a hard problem because the voice consists of a bunch of fundamentals, plus partials, plus sidebands. Any chance in pitch to
    any of them, or any change in their levels, would cause a timbral
    change. Would it be audible? Maybe, depending. I can drop a deep notch
    into a vocal track at one frequency, and you wouldn't notice it, and I
    can drop it somewhere else and it would be obvious.

    Back in the seventies, Stockham took digital notch filters to a Caruso recording, with the intention of notching out all of the horn resonances
    that were part of the recording process. But the end result didn't sound
    like Caruso... much of the depth of his voice was gone. Was this because natural vocal resonances were getting notched out? Or was it because
    Caruso's voice wasn't really as complex and thick and people remembered it
    as being? Folks have been arguing over that one for decades.
    --scott
    --
    "C'est un Nagra. C'est suisse, et tres, tres precis."

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Orlando Enrique Fiol@21:1/5 to [email protected] on Sun Dec 3 15:23:16 2023
    In article <uki2ir$9rd$[email protected]>, [email protected] writes:
    For voice that's a hard problem because the voice consists of a bunch of >fundamentals, plus partials, plus sidebands. Any chance in pitch to
    any of them, or any change in their levels, would cause a timbral
    change. Would it be audible? Maybe, depending.

    Plus, the vocal timbre would need to be distinguishable from other instrumental timbres.

    I can drop a deep notch into a vocal track at one frequency, and you wouldn't notice it, and I
    can drop it somewhere else and it would be obvious.

    This only takes vowels into account, whose frequencies could be easily measured.

    Back in the seventies, Stockham took digital notch filters to a Caruso >recording, with the intention of notching out all of the horn resonances
    that were part of the recording process. But the end result didn't sound >like Caruso... much of the depth of his voice was gone. Was this because >natural vocal resonances were getting notched out? Or was it because >Caruso's voice wasn't really as complex and thick and people remembered it
    as being? Folks have been arguing over that one for decades.

    Considering how much horn resonances were both highly colored and woefully incomplete in terms of frequency ranges, Caruso's voice probably sounded little like his recorded representations. Only live recordings using contemporary proaudio equipment would settle this question accurately/.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)