Log in

View Full Version : Fixing/regenerating clipped audio?


dansrfe
6th August 2013, 06:36
I recently bought a bluray for an old movie and nearly all of the 6 channels have been run through a low-pass filter, center channel especially so. Now, I realize that clipped audio cannot be restored however is there a way to algorithmically generate the clipped high frequencies to a certain degree to mitigate the effects of the audio sounding "covered up"?

Thanks in advance.

Ghitulescu
6th August 2013, 09:30
I assume the reason to do this was that the lower part of the spectrum contained hum, a common issue back then. Also the old movies did not have HD Audio :), so adding something that wasn't there in the first place requires a bit of imagination.
Sometimes, the DVD issues of the same movie might help.

pandy
6th August 2013, 10:08
I recently bought a bluray for an old movie and nearly all of the 6 channels have been run through a low-pass filter, center channel especially so. Now, I realize that clipped audio cannot be restored however is there a way to algorithmically generate the clipped high frequencies to a certain degree to mitigate the effects of the audio sounding "covered up"?

Thanks in advance.

I don't understand - clipped means level is cut (amplitude was to high), low pass filter (lack of high frequencies) is something else than clipped.

For really clipped audio not many things can be done, Audacity offer some limited processing with audio restoration tools.
For low pass audio technique so called Spectral Band Replication can be used to add artificially generated high frequencies that will improve perceived audio.
FFDshow is equipped with such filter (named "Crystality") - if your audio editor support direct show audio filters then ffdshow can be used however it can be unstable and will work more or less correctly only with one sampling frequency (44100) so sample rate conversion is recommended.

dansrfe
6th August 2013, 22:45
It's not a really old movie actually (2005). The thing is that the bluray is a remastered version of a film that was done by REM. It has two audio tracks, German and Hindi in dts-hd ma. Now, the German audio is perfect as well as the songs (which they kept intact in the hindi) however, the hindi track of the film sounds like it did with the original DVD release. Deep hum sound. Almost like I'm hearing it from another room. I'm not sure if that's clipping or something else.

Although it's not a topic of this thread, I'm quite interested to know how REM managed to work with the original hindi track and fix it to near pristine audio fidelity for use in the German track. Did they have a different source to work from altogether? My main goal here is to replace the German track's center channel with the Hindi track's center channel yet use all of the channels in the song portions of the German track.

Two issues with this:

1. If I cannot manage to get the Hindi center track to match the audible frequencies of the German track's side and back channels then the final mix will sound odd and the whole process will be moot.

2. I'll have to find the parts of the audio in which the dialog delivered has an echo effect which permeates through the side and back channels. In those parts I'll have to use the affected channels (probably all) from the hindi track and apply the same fix to those channels as done to the center channel.

It might be time to start learning German. :o

dansrfe
8th August 2013, 08:55
I don't understand - clipped means level is cut (amplitude was to high), low pass filter (lack of high frequencies) is something else than clipped.

For really clipped audio not many things can be done, Audacity offer some limited processing with audio restoration tools.
For low pass audio technique so called Spectral Band Replication can be used to add artificially generated high frequencies that will improve perceived audio.
FFDshow is equipped with such filter (named "Crystality") - if your audio editor support direct show audio filters then ffdshow can be used however it can be unstable and will work more or less correctly only with one sampling frequency (44100) so sample rate conversion is recommended.

Yeah, I'm not really sure of the specific term to accurately describe the issue.

I was looking at trying Izotope RX 2. Would Spectral Repair, Declipper, or the Hum Remover be of use in this situation? Should I look into trying some other software? Will Audacity give provide optimal results in your opinion? To be honest, I'm not too sure where to begin with the software.

Also, would uploading a small cut of the dialog and/or song portion from both german and hindi tracks help identify the problem? With or without the echo parts?

Thanks.

Ghitulescu
8th August 2013, 10:40
German tracks are always reworked, so they sound better even than the originals (it includes Hollywood here, not only Bollywood).

dansrfe
8th August 2013, 12:01
By reworked do you mean that they start from a different source provided by the film studio altogether or they have specialized audio engineers that actually fix the audio from the original source?

Brazil2
8th August 2013, 14:45
FFDshow is equipped with such filter (named "Crystality") - if your audio editor support direct show audio filters then ffdshow can be used however it can be unstable and will work more or less correctly only with one sampling frequency (44100) so sample rate conversion is recommended.
Still, ffdshow is a great piece of software, a jewel, and it's a pity no ones wants to take over its development :(

dansrfe
8th August 2013, 23:38
Here's a 1min cut of the same time frame from the center channels of both language tracks: https://www.dropbox.com/s/lglh21tsa1k57pj/comparison.zip. Honestly, I'm not sure where to begin. The german audio sounds like it was recorded and mastered entirely differently.

raffriff42
9th August 2013, 03:22
It's got what sounds to me like moderate intermodulation distortion (http://www.triggertone.com/term/Clash), spread across all frequencies. Probably a bad optical sound track. Can't filter that stuff out. Exiters only make it sound worse. Best result for me was with a 5kHz low pass, followed by a treble boost hinging at 1.5kHz.

dansrfe
16th August 2013, 18:00
Hey raffriff42, which program did you use to run the audio through a low pass/treble boost? Audition or something else? Thanks!

raffriff42
16th August 2013, 18:45
Audacity :) IIRC, I did something like:
Low Pass, 5000 Hz, 24 dB/8ve, followed by
Equalizer, with all sliders flat up to 1 kHz, gradually increasing to about +8 dB at 5000 Hz (in hindsight, should have simply used Bass and Treble)