Audio Restoration - AWFUL file

bedbugs

If anyone can help I'd really appreciate it. A friend has done an interview and the samplerate is at 22.05khz. Nasty.

I've tried using Soundforge and iZotope Rx. If anyone has any genius ideas, could you let me know asap? This yoke is urgent.

thanks

c

ogy

try a combination of high pass and low pass filtering, might make it sound telephony but hopefully would get rid of some nasty artifacts

bedbugs

Thanks for the reply, but I've tried that.

Neurojazz

Dynamic EQ? - that's awesome for removing hard to catch problem frequencies.

madtheory

Post it up and let us hear it, far easier to diagnose.

Izotope Rx is for noise removal. Is it noisy?

22050Hz should actually be perfectly good for speech! If it's lacking clarity, try a wide 4k boost.

jimi_t

madtheory wrote: »

22050Hz should actually be perfectly good for speech! If it's lacking clarity, try a wide 4k boost.

I'd say the real problem here is that it was recorded with a condenser or some other unsuitable mic. 32 band EQ combined with noise removal from Isotope might be the best course of action - its never going to sound perfect though.

PaulBrewer

bedbugs wrote: »

If anyone can help I'd really appreciate it. A friend has done an interview and the samplerate is at 22.05khz. Nasty.

I've tried using Soundforge and iZotope Rx. If anyone has any genius ideas, could you let me know asap? This yoke is urgent.

thanks

c

If it doesn't contain private info, post it up till we have a listen.

woodsdenis

bedbugs wrote: »

If anyone can help I'd really appreciate it. A friend has done an interview and the samplerate is at 22.05khz. Nasty.

I've tried using Soundforge and iZotope Rx. If anyone has any genius ideas, could you let me know asap? This yoke is urgent.

thanks

c

http://www.proaudiodsp.com/

This will fix almost anything , apart from clicks and pops etc. Demos available. Watch the vids to see if it is applicable.

madtheory

jimi_t wrote: »

I'd say the real problem here is that it was recorded with a condenser or some other unsuitable mic.

You could well be right. But we have no idea til we hear it surely?

jimi_t wrote: »

32 band EQ combined with noise removal from Isotope might be the best course of action - its never going to sound perfect though.

Why you you limit yourself with a graphic? Surely several bands of fully parametric eq would give more control, and less undesirable artefacts?

bedbugs

Thanks guys. The file was from a dictaphone, from a journalist who needed it sorted. The distortion was ridiculous. Even with all tries of noise reduction and EQing, it was impossible to make out even one word of the interview.

A lost cause.

jimi_t

madtheory wrote: »

You could well be right. But we have no idea til we hear it surely?

Dictaphone/Condenser. I've had the same problem before.

Why you you limit yourself with a graphic? Surely several bands of fully parametric eq would give more control, and less undesirable artefacts?

Fixed band first, parametric after - otherwise its almost impossible to narrow the problem areas out; I find it counter-intuitive any other way.

madtheory

Ah. Clipping distortion is almost impossible to deal with. This is a shortcoming in education I think. Monitoring off tape is vital! One should at least do a test recording before the real interview gets going, to ensure all is well.

jimi_t wrote: »

Dictaphone/Condenser. I've had the same problem before.

I see! Which models? How does this happen? Is there no user control over level?

jimi_t wrote: »

Fixed band first, parametric after - otherwise its almost impossible to narrow the problem areas out; I find it counter-intuitive any other way.

So you're using the graphic just to locate the problem frequencies? Maybe a spectrum analyzer would be more useful? Analogue 31 bands often cause more problems than they solve. In the analogue and digital domain, you're very constrained with the frequency and bandwidth, whereas the parametric gives total flexibility. So you can treat only the problem areas, leaving the wanted audio intact.

kfoltman

bedbugs wrote: »

Thanks guys. The file was from a dictaphone, from a journalist who needed it sorted. The distortion was ridiculous. Even with all tries of noise reduction and EQing, it was impossible to make out even one word of the interview.

The distortion? In literal sense, like, it was recorded so hot that the signal was clipped? I think CoolEdit Pro used to have a filter that attempted to correct the hard-clips, but there are probably better tools now. In that case, EQ or noise reduction wouldn't help much. I doubt anything can be done, as the clipped part of the signal are basically lost (straight-ish line in the waveform). Still, I thought that typically dictaphones have a gain control circuit (compressor/limiter basically) to avoid stuff like that.

Or distortion as in generally poor signal quality?

bedbugs

kfoltman wrote: »

Or distortion as in generally poor signal quality?

yep

kfoltman

By the way, if you need a solid-ish argument against blaming sample rate for this particular fiasco: the old telephone networks used 8 kHz sample rate (64kbps bitrate, 8 bits/sample with sort of logarithmic mapping to handle quiet signals nicely), and the speech was not *totally* illegible.

Not that it matters now - just a bit of trivia.