Tim Gillett wrote:I doubt it's possible from a limited bandwidth recording to "synthesise" the actual missing overtones of that voice.
It is possible, with Izotope Spectral Repair.
Great, any chance of an audio example please?
You can experiment for yourself with the 30-day free trial of RX8.
The Spectral Recovery Module works a bit like an aural exciter in synthesising upper harmonics (above 4kHz) from the audio content, but in a way which is specifically optimised for repairing speech, recognising the difference between vowels and sibilants.
It can also fill in spectral holes within the audio (below 4kHz) created by lossy compression using the same core technology.
The module has a Learn function to analyse the content and determine appropriate settings, but there are also five user controls for preferential optimisation. (Source cut-off frequency, amount, vowel/sibilant balance, smoothing, and spectral patching on-off).
I found it pleasingly effective and natural, and a useful new tool to have in the armoury.
Perhaps so but what was the reference?
Actually my 30 day trial of RX8 expired months ago. I still use it for evaluation but the 30 second limit now makes such evaluations of its tools much more difficult.
Hugh Robjohns wrote:As a business expense I'd say it can be justified, and it's good to stay current... but for a hobbiest I think it really depends on whether the new stuff is directly relevant to your specific needs.
Thanks for that, Hugh
I use various bits of it on almost everything I do....but most of the new features (barring possibly Spectral Repair) would be relatively infrequent at this moment in time...I've been through 5-7 Advanced but normally end up buying just at the wrong time just before the next major update comes out!
Last edited by Aural Reject on Wed Dec 09, 2020 1:17 pm, edited 2 times in total.
Hugh Robjohns wrote:And the W&F module is okay and handy for simple jobs, but a little limited for real-world tricky stuff...
I'm probably a little more enthusiastic about this than Hugh but then I don't have access to Capstan or Respeed to compare it with. It works well on cyclical variations like an off centre record or a cassette recorded with a dirty capstan. However, it doesn't seem to handle incidents like someone brushing against a tape flange where the recorded audio could speed up by a few semitones for a second or two before returning to normal. I also found it seemed to work better when you tell it to reference the audio to a known pitch. W&F was the main reason I upgraded to Advanced this time round.
Last edited by Hugh Robjohns on Wed Dec 09, 2020 3:04 pm, edited 2 times in total.
A few posts back a few users suggested converting the stereo tracks that are identical left and and right to a mono track. What is the downside of leaving the track stereo? CPU efficiency?
Thanks everyone, this forum is really helping quarantine be more tolerable!
Yes, and storage, the files will be half the size. I just think it's neater though and in an uncharacteristic attack of OCD I would reduce any 'dual mono' tracks to mono..
