not sure how to frame question - "sharpen old audio" ?

[hydra3333]

Some old moves, eg black and white ones, tend to have what appears to be somewhat soft or muffled audio.

I currently use ffmpeg to detect volume levels and adjust ... however I wonder if there is such a thing as "sharpening audio" in an equivalent way to sharpening video ? Dialogue, not music I'm not even sure of what terminology to use.

Advice would be appreciated.

[JVRaines]

Just for dialogue? I would look at equalization to lower muddy resonances, harmonic exciter to crisp it up, and multiband compression to tighten the dynamic range if necessary.

[hydra3333]

Thank you. As a first step then, would this feature be available in ffmepg ?
https://en.wikipedia.org/wiki/Exciter_(effect)

[smrpix]

You want to be able to hear what you're doing and make adjustments. Try pulling your tracks into a DAW, even something simple and free like Audacity. Start with the equalizer and give a boost in the 1-2k range. Tweak from there.

Be aware that old optical tracks had a limited frequency range, so you'll never completely reconstruct missing highs or lows.

[pandy]

There is Crystality (spectral band replication) filter in ffdshow - you can use it (beware - it is extremely unstable, also some delay is added before audio so it need to be trimmed after applying filter).

[hello_hello]

As smrpix suggested, Audacity tends to be useful for this sort of thing. For Lo-Fi, muffled audio, check to make sure there's not a lot of background noise or low hum, If there is, removing as much as possible can increase the clarity substantially and reduce the amount of EQind and or compression etc required. Audacity has a noise removal filter. You need to select a section where there's noting but noise, get the filter to analyse it, then apply the noise removal to the entire audio. If there's a lot of noise and you try to remove too much it'll start to adversely effect the remaining audio, so it pays to adjust the "strength" of noise removal and experiment a little.

[hydra3333]

OK thank you. I have used Audacity in the past and it's a really handy tool.

I have a batch commandfile driven workflow to transcode multiple files, "fire and forget", so ffmpeg is very handy using somewhat generic settings for old sources.

In this case I suppose I will need to look into "Crystality (spectral band replication) filter in ffmpeg and whether the delay is constant. I had wondered if SOX or similar could excite me but noone mentioned it.

[Cornucopia]

99% of correction/restoration/enhancement work cannot be done using stock/boilerplate settings. You'll have to manually set up each plugin/filter & adjust each setting.
Strongly recommend you use a DAW.


Scott

[hydra3333]

That is all good advice. OK.

With a larger than smaller number of old sources which will be watched very infrequently, it's a time vs benefit consideration.

With that in mind, a view "leave it alone or any tiny automated improvement would be welcome" would be a pragmatic approach.

[hydra3333]

There is Crystality (spectral band replication) filter in ffdshow

Would that be the same as https://ffmpeg.org/ffmpeg-all.html#crystalizer ? It doesn't sound the same: "Simple algorithm to expand audio dynamic range."

Or maybe this https://www.ffmpeg.org/ffmpeg-codecs.html#Options-7 as an option of libfdk_aac "eld_sbr Enable SBR (Spectral Band Replication) for ELD if set to 1, disabled if set to 0."

They're the only hits I could see.

[hydra3333]

I noticed this https://sourceforge.net/projects/crystality/ where its readme says

Crystality audio DSP Plugin README file
Plugin for realtime 'remastering' sound leaned by lossy compression
Bandwidth Extender (aka Exciter), Harmonic Booster and 3D Echo.

Which is what I think is suggested above and what I think I need

I wonder, would this be the same as ffmpeg's Crystalizer - the ffmpeg doco doesn't sound the same: "Simple algorithm to expand audio dynamic range" ?
https://ffmpeg.org/ffmpeg-all.html#crystalizer