• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Thimeo Perfect Declipper

Patrick1958

Senior Member
Forum Donor
Joined
Jun 28, 2018
Messages
498
Likes
412
Location
Belgium
The Themeo Perfect Declipper is a tool used to correct (severe) clipping in an audio file. Someone mentioned it in a topic before, it might be useful to see some measurements and screen captures of the tool's results.

The Song used : Creedence Clearwater Revival : Have You Ever Seen The Rain 24/96 download.
The entire album shows heavy processing (loudness war?) this song from the album as an example clearly showing clipping in Audacity
Waveform Before.jpg


Below the same song run trough Perfect Clipper with standard settings :
Waveform After.jpg


Clearly much better.
DR improved also.
Original file :
DR Before.jpg


After Declipping :
DR After.jpg


Dynamic range improved by 2 points.
Tested this with the entire album and on two songs even an improvement of 3 points.
I leave it in the middle how much of these improvements are audible. Decide for yourself with your own ears ;)
 

Veri

Master Contributor
Joined
Feb 6, 2018
Messages
9,596
Likes
12,036
The "dynamic range" is improved, because the algorithm fixes problematic peaks, but does it fix what's already messed up in the sound..?

I'm doubtful.
 

Snarfie

Major Contributor
Forum Donor
Joined
Apr 30, 2018
Messages
1,169
Likes
926
Location
Netherlands
The Themeo Perfect Declipper is a tool used to correct (severe) clipping in an audio file. Someone mentioned it in a topic before, it might be useful to see some measurements and screen captures of the tool's results.

The Song used : Creedence Clearwater Revival : Have You Ever Seen The Rain 24/96 download.
The entire album shows heavy processing (loudness war?) this song from the album as an example clearly showing clipping in Audacity
View attachment 19674

Below the same song run trough Perfect Clipper with standard settings :
View attachment 19675

Clearly much better.
DR improved also.
Original file :
View attachment 19676

After Declipping :
View attachment 19677

Dynamic range improved by 2 points.
Tested this with the entire album and on two songs even an improvement of 3 points.
I leave it in the middle how much of these improvements are audible. Decide for yourself with your own ears ;)
Did some recordings a few days ago with pictures an audio for download. https://www.audiosciencereview.com/...ighly-compressed-recordings.5923/#post-132926
 

pos

Addicted to Fun and Learning
Forum Donor
Joined
Feb 13, 2018
Messages
572
Likes
718
I just tested that tool on different versions(*) of Norah Jones' utterly clipped "Don't Know Why" title, and I must say I am really impressed with how effective it is!
Of course it does nothing for the numerous digital clicks that clutter this song, but it does audibly reduce clipping artifacts!

You can judge for yourself with these 6 sec excerpts here: https://www.dropbox.com/s/q4ebdo4q16899sw/Declipper test - Norah Jones.zip?dl=1
(most notable clipping passage is on the word "on")

All version are normalized to the same RMS volume within around 0.1dB (gain in the filename, unchanged between original and declipped versions).
Looking at the RMS values of theses excerpts, it appears the 2008 version is 2dB louder than the 2002 one, and the hires version is 0.8dB quieter (and inverted in polarity?!).

(*) I cannot assert for the origin of the 2008 (mp3) and hires (96/24) versions, but I ripped the 2002 one myself from the original CD album.
 
Last edited:

Dro

Active Member
Joined
Apr 8, 2018
Messages
221
Likes
205
It works well enough on simple straight clipping or very hard limiting to just use it as a DSP for live processing. If you want to get into more serious restoration, especially clipping that is tilted, I recommend getting a proper VST host and then running a little DSP chain.

I use this chain:
Stereotool (Declipper)
RX6 De-click (removes clicking and crackling artifacts that might be left over)
Ozone 8 Elements (Compressor)
Stereotool (Natural Dynamics, True Bass)
Ozone 8 Elements (EQ, Compressor)

The izotope plugins are both very user friendly. The point of this chain is consistent performance, even under very bad circumstances. The declipper output might introduce a few very high peaks or strong channel imbalance. The compressor after the declipper will help with that a lot, feeding a better signal for Natural Dynamics to process. True Bass helps with getting some extra punch below 100 Hz, which often seems to be EQed down to allow for higher loudness. Quite often the 100 to 200 Hz range seems to be a bit boosted to make up for that, thickening the sound, so I tend to EQ that down in the second instance of Ozone.

I only gain around 2 dB dynamic range from the declipping and dynamics processing each, for a total between 3 and 5 dB. This may not sound as good as just running default settings and letting it go up to +12 dB, but in practice a consistent dynamic range increase around 4 dB is much preferable to me over freak peaks, channel imbalances and false negatives in the declipper stage.
 

exaudio

Member
Forum Donor
Joined
Jun 11, 2018
Messages
46
Likes
75
I've been curious about how well these declipper tools work to "restore" the information lost to clipping. I took the clips posted by @pos and loaded them into Audacity to compare the before and after tracks. I wouldn't claim to be able to hear a difference, but to my eye it looks like the algorithm does a reasonable job of "guessing" what the samples might have looked like had they not been lost to clipping.

The top line is the left channel of the clipped track. The middle line is the left track declipped and the bottom is the difference between the two (which I think is what the declipper must have added).

declipped.png
 

Dro

Active Member
Joined
Apr 8, 2018
Messages
221
Likes
205
Looks like a few false positives here. But those can usually taken care of by adapting the settings to the specific track.
 

RayDunzl

Grand Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
13,204
Likes
16,985
Location
Riverview FL
I've been curious about how well these declipper tools work to "restore" the information lost to clipping.


Howsabout you try this:

Take an unclipped track segment.

Increase the amplitude to create clipping.

Declip that.

Compare to the original.
 

pos

Addicted to Fun and Learning
Forum Donor
Joined
Feb 13, 2018
Messages
572
Likes
718
I took the clips posted by @pos and loaded them into Audacity to compare the before and after tracks. I wouldn't claim to be able to hear a difference, but to my eye it looks like the algorithm does a reasonable job of "guessing" what the samples might have looked like had they not been lost to clipping.
The difference is pretty obvious if you listen carefully with descent headphones or speakers, especially with the 2008 compilation version.

I made the test @RayDunzl describes, and the declipped version was different than the original. What this declipper does is not restore the original, but remove audible artifacts by replacing them with inaudible ones ;)
It is not bitperfect by any mean, but it certainly does its job when used on audibly clipped contents like the excerpts above.
 

Dro

Active Member
Joined
Apr 8, 2018
Messages
221
Likes
205
The default use of Stereotool is real time processing.
 

Dro

Active Member
Joined
Apr 8, 2018
Messages
221
Likes
205
Now, while live processing is quite capable for most of the less affected music, here is an example of where the results are not very good.
advanced processing.png

At the top you can find the original file. In the center, bright, the output of just letting Stereotool run with Declipper, Natural Dynamics and True Bass on, using default values. At the bottom is the output of a more complex DSP chain, also using Declipper, Natural Dynamics and True Bass.

RMS is matched for all tracks. The original peaks at around -6 dBFS. The simple processing result has been clipped at 0 dBFS so we get a better view of the audio. The output of the DSP chain at the bottom peaks at -1.4 dBFS.

You can see that:
  • Simple processing on paper has the highest dynamic range
  • Simple processing also potentially introduces a lot of channel imbalance for peaks
  • Simple processing has very bad clipping detection for tilted clipping, making its declipping restoration inferior and pratical dynamic range is inferior to the complex processing
 

ryanmh1

Member
Joined
Jan 11, 2019
Messages
93
Likes
117
Here's a link to another post of mine that has a few album links that someone processed using Stereo Tool. https://www.audiosciencereview.com/...d-why-is-it-happening.3392/page-7#post-152646. This difference this tool makes is phenomenal, in my opinion. Depending on your listening tastes, easily a better way to spend $100 than on pretty much anything else. It seems to magically recover details that shouldn't even be there. But somehow, they are.

I haven't read through the whole thread, but I'm guess that this probably works best with albums that were brutally hard clipped, which probably makes it fairly easy for the software to make very educated guess about what ought to be there. I think it's fair to say that whatever it does put there, the final product is better than the original. Sort of like a map that is missing a piece. You can take a fairly good guess at what should be there based on the surroundings, and even if it isn't perfect, it is still a lot better than just leaving a hole.
 

Tool

Active Member
Joined
Jun 29, 2019
Messages
145
Likes
92
I'm using Perfect Declipper as a vst for Adobe audition 6.0 64 bits, and I'm really impressed. The files I'm sending are logs of Dynamic Range Meter 1.1.1 from foobar. Try to compare it to http://dr.loudness-war.info:

These are files for copule of albums there were compressed around Dr7 and Less, and check the results. Sounds far more better now. I can send a sample sound if someone is intrested.
 

Attachments

  • foo_dr.txt
    1.5 KB · Views: 148
  • (2013) Lightning Bolt_log.txt
    1.5 KB · Views: 170
  • foo_dr.txt
    1.5 KB · Views: 136
  • foo_dr.txt
    1.2 KB · Views: 118

Tool

Active Member
Joined
Jun 29, 2019
Messages
145
Likes
92
Unfortunatly Loudness war doesn;t meess up only with pop music like metal, punk rock or alternative but Jazz as well. To anyone who does it keep in mind this picture:
And please stop doing this...
 

Attachments

  • GTxV5eJ.jpg
    GTxV5eJ.jpg
    58.6 KB · Views: 353
OP
Patrick1958

Patrick1958

Senior Member
Forum Donor
Joined
Jun 28, 2018
Messages
498
Likes
412
Location
Belgium
I'm using Perfect Declipper as a vst for Adobe audition 6.0 64 bits, and I'm really impressed. The files I'm sending are logs of Dynamic Range Meter 1.1.1 from foobar. Try to compare it to http://dr.loudness-war.info:

These are files for copule of albums there were compressed around Dr7 and Less, and check the results. Sounds far more better now. I can send a sample sound if someone is intrested.
Impressive numbers. Care to share your thimeo settings?
 

Dro

Active Member
Joined
Apr 8, 2018
Messages
221
Likes
205
It is very easy to get high DR numbers with little audible differences, depending on the settings. DR after declipping tends to be as reliable as a random vinyl rip.
 

Tool

Active Member
Joined
Jun 29, 2019
Messages
145
Likes
92
Let me show you what is a difference between Vinyl rip measurement and cd declipped. System Of A Down (2005) Mesmerize. Just for you to know, I leave every Vinyl Rip as it is, Only Cd's are corrupted and that's unacceptable.
 

Attachments

  • Messmerize cd declipped.txt
    1.4 KB · Views: 178
  • Vinyl Messmerize.txt
    1.4 KB · Views: 149

Tool

Active Member
Joined
Jun 29, 2019
Messages
145
Likes
92
Impressive numbers. Care to share your thimeo settings?

If you ask me about settings I don't change it for cd's rip. So I use only Declipper and Natural Dynamics, but when I need to retrieve sound from mp3 its more complicated.
 

Tool

Active Member
Joined
Jun 29, 2019
Messages
145
Likes
92
Let me show you two exemples what Loudness War does to the Music:
This is 2 albums of Ill Nino band albums original cd rip and after declippiner :
 

Attachments

  • (2010) Dead New World Original cd.txt
    1.5 KB · Views: 140
  • Capture1.JPG
    Capture1.JPG
    551 KB · Views: 187
  • (2012) Epidemia Original Cd.txt
    1.3 KB · Views: 133
  • Capture 2.JPG
    Capture 2.JPG
    551 KB · Views: 260
  • foo_dr.txt
    1.5 KB · Views: 104
  • Capture.JPG
    Capture.JPG
    724.1 KB · Views: 205
  • foo_dr.txt
    1.3 KB · Views: 113
  • Capture 2.JPG
    Capture 2.JPG
    732.1 KB · Views: 235
Top Bottom