Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Declipping (Read 13528 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Declipping

Hello,

can anybody recommend better declipping plugin for Audacity or other software than Audacity's ClipFix?

Preferable open source/freeware. I have heard about Re-Life but read here that it does not do in fact anything significant with the sound.

Thanks

Jan

Re: Declipping

Reply #1 –
Declipping is a false economy at best. It tries to recreate information based on basically no data. While some algorithms do look at surrounding samples, it's still guesswork. It may be fine if you only have very few clipping samples, or the input signal is relatively simple. But it's hopeless if you're trying to correct a completely flat-topped waveform in the middle of a complex musical signal.

The only time I've actually seen clipping being completely reversed was on an MP3, where the global gain was erroneously set too high in the file, either because of a bad encode or misuse of MP3Gain. The waveform was completely clipped in Audacity, but fixing the global gain fields with MP3Gain at reasonable settings undid the clipping.

Re: Declipping

Reply #2 –
I know that declipping cannot create new original waveforms from nothing. But if I have lightly clipped (and thus to some extent distorted) tracks, when I apply ClipFix, I like the result better than the source. Or should I stay with clipped version at all costs?

Re: Declipping

Reply #5 –
Quote from: jumpingjackflash5 on
when I apply ClipFix, I like the result better than the source.
In terms of sound quality and based on properly controlled (time-synched/level-matched/etc.) double blind testing, no doubt.  ::)

...if not, then relife makes some pretty snazzy looking waveforms and TTDR ratings might even turn from red to green. :D

Re: Declipping

Reply #6 –
I feel irony in that.

Re-Life as been discussed here and classified as not working or only phase-changing. https://hydrogenaud.io/index.php/topic,80679.0.html

If declipping is totally hopeless I can let it be.

If it can be done better than with ClipFix or ReLife then please suggest how.


Re: Declipping

Reply #7 –
Quote from: greynol on
Quote from: jumpingjackflash5 on
when I apply ClipFix, I like the result better than the source.
In terms of sound quality and based on properly controlled (time-synched/level-matched/etc.) double blind testing[?]
Quote from: jumpingjackflash5 on
I feel irony in that.
I'll take that as a "no" then.

Re: Declipping

Reply #8 –
OK then. For now I'll use Audacity's ClipFix, since nothing better is around, even if I know that does not do any "magic" for this purpose.

Re: Declipping

Reply #9 –
You have me curious as to whether it does anything worthwhile other than produce prettier waveforms, though I shouldn't be casting any more doubt than what has been done already. :)

I think Monty has some declipping algorithm somewhere.

Re: Declipping

Reply #10 –
Yes i have read about it but it is supposed to run on linux (unix) only which unfortunately do not have at all.

Re: Declipping

Reply #11 –
Quote from: greynol on
You have me curious as to whether it does anything worthwhile other than produce prettier waveforms

That's pretty much what I concluded after trying it. The waveforms look great and dynamic, but I didn't hear any improvement, it still sounded distorted.

Re: Declipping

Reply #12 –
Yes they do look great that is why I first tried Re-life until I read posts here. I can ABX audacity Clipfix versions(although in other abx tests I usually fail), but I think it is mainly because of volume changes done before Clipfix (to make room for waveforms).

Just to add up, the tracks I tried to declip were only lightly clipped and they had intentionally been distorted (chicago blues style, unfortunately too much), with dynamic range 7-9, afterwards the dynamic range was 10-11 dependent on the extensity of changes.  

Re: Declipping

Reply #13 –
The DR measurement is useless, you're creating fake dynamic range.

The biggest issue is that the lost content simply isn't predictable by an algorithm that only understands waveforms, not complex musical content. The best it can do is interpolation based on preceding and following sample values, but it will always be a mathematical guess, not a musical one.

I found that the resulting "declipped" version sounded just as distorted to my ears as the original clipped version.

And if you overdo it, you get "pumping" artifacts, like what happens when well-meaning amateurs with sound editors try to expand the dynamic range of overly-compressed albums. The DR rating and ReplayGain values look great, but the end result sounds terrible.

Re: Declipping

Reply #14 –
Thanks for clarifying. So the new DR is only the measurement of the "dimensions" of new waveforms, not the sound itself.

I do not overdo it (I reduce volume by -3/-5 dB and apply ClipFix at 95 per cent), I can do further testing at comparable volume levels, if ClipFix brings anything more (audibly - in practice) than Re-Life.

But theoretically, even if the new waveforms done by ClipFix are only "mathematical", the results could be in some sense  better than just clipped version with decreased volume-deamplified (?).

Re: Declipping

Reply #15 –
If you can hear the differences, then whether it is better or not is just your personal preference.

If you can't hear the differences, why bother?

These kinds of questions clearly show that your mind didn't change a bit since your first post in January.

Re: Declipping

Reply #16 –
I dont wanna get personal but I learned a lot of knowledge since January.

As I wrote before I can hear the differences with ClipFix easily in ABX, but I wonder whether they are caused just by volume changes (required before clipfix to make it work), or also by new waveforms and their character, or by other factors that come into processing.

That is why I asked whether better software than ClipFix is available, since ReLife has been discussed here as not working at all.

Re: Declipping

Reply #17 –
The foobar ABX tool already has an option to match the gain of two files.

Re: Declipping

Reply #18 –
Thanks for a tip! I will try it.

Re: Declipping

Reply #19 –
Quote from: KozmoNaut on
Declipping is a false economy at best. It tries to recreate information based on basically no data. While some algorithms do look at surrounding samples, it's still guesswork. It may be fine if you only have very few clipping samples, or the input signal is relatively simple. But it's hopeless if you're trying to correct a completely flat-topped waveform in the middle of a complex musical signal.

I've done some ABXing (long ago when my ears were still at least average) and found that modest amounts of clipping can be inaudible - it is highly dependent on the program material.  Pure low frequency tones are among the program material that makes even modest clipping most audible.

Re: Declipping

Reply #20 –
Quote from: bennetng on
The foobar ABX tool already has an option to match the gain of two files.
I doubt level matching with replaygain will be good enough in this situation.

Re: Declipping

Reply #21 –
If clipfix software only tries to paint nice waveforms it may ba a good idea to try some decrackle/declick software at the same spots and listen if the distortion betters.
Is troll-adiposity coming from feederism?
With 24bit music you can listen to silence much louder!

Re: Declipping

Reply #22 –
Incorrectly stylizing it as "ClipFix" could have been misleading. Clip Fix is a function that appears to be built into Audacity:
http://manual.audacityteam.org/man/clip_fix.html

If the function requires a manual level change in order to provide headroom (per the link) then the same level change should be applied to the original source when making qualitative comparisons.

Re: Declipping

Reply #23 –
Apart from level-matching techniques, how about deliberately clip/distort some files in different ways and use different declippers to restore them and compare with the original files? At least we have a reference to what an undistorted file sound like.

[edit]Someone tried this method in previous discussions already :P

Re: Declipping

Reply #24 –
While it has been discussed before, it's worth mentioning that ABX is for demonstrating *differences*.  by itself is not the right tool for objectively determining *preferences*, as A and B are known throughout the test. If B is the prettier waveform then this knowledge can serve as a bias.