Individual clicks/pops highlighted in whole wave view

[not]

It ain't easy, I know, I know, but it would be outSTANDing if we could apply a function/filter to a waveform in edit view that a) highlighted potential clicks/pops, AND b) added the potentials to a cue-type list, so I can just jump from one potential problem to the next. Spectral view would be optimal, but sample view would handle it, too.

I know there are a million reasons why not, but after autofixing clicks with (insert favourite algorithm here), I have to listen to every millisecond of audio to identify the 20 or 30 clicks that can't be taken out automagically without mangling the underlying audio.

Hell, I know what a click looks like, and AA shows the damn things so clearly...sob...

This is all presented from the perspective of a recording recovery job.

Sigh. Fingers crossed.

[not]

I know there are a million reasons why not, but after autofixing clicks with (insert favourite algorithm here), I have to listen to every millisecond of audio to identify the 20 or 30 clicks that can't be taken out automagically without mangling the underlying audio.

Hell, I know what a click looks like, and AA shows the damn things so clearly...sob...

Whilst I understand exactly which bit of utopia you are after, there is a slight logical problem... since Audition clearly can't identify all clicks (which I'd have to agree about), then it couldn't identify them to display them as such either. My experience of clicks, pops, thumps, bumps and other things that you can usefully remove with 'fill single click now' is that they don't all have the same risetime characteristics at all, and in fact vary quite widely. And as such, can't be unambiguously identified. There are some percussion tracks  I've treated where it's been virtually impossible to treat most of the obvious clicks anything other than manually - in order to identify them, the software would not only need to be able to differentiate clicks, but also have to have a knowledge of the piece itself, and it clearly can't have that.

It's a nice idea, but clicks are not 100% predictable, and so unless there is a major breakthrough in technology that understands what things should sound like, I don't think it's going to happen. Mind you, having an option to display (within click removal) what the program is actually going to remove from a file would be useful as a setup tool - although using this on the entire file in advance would inevitably increase the processing time significantly. And, having all the clicks that Audition could potentially identify on its own in the cue list is going to generate a list so long with some material that just this 'feature' on its own could cause you to lose the will to live, I fear...  

Ain't it wonderful how we can listen to a track and identify unwanted sounds instantly, but we still can't write software to do it?
Makes me feel all warm and tingly and human

While the objections expressed are true, there is a program that does much as you requested. WaveRepair has a fair number of options about what to hunt for. It marks each instant that it finds and allows you to jump from one to the next. Thus, you can have the program search under one criteria, you review and fix as necessary, then search under the next, etc. Its click fix options are excellent.

The downside of the program is that it only works in 16 bit, 44.1kHz or 48kHz. Thus you probably want to put off the final click clean-up until after everything that needs 32 bit is finished. Since manual click repair is very local, there isn't any 16 bit penalty for the click fixing itself, but you probably don't want to change from 32 bit to 16 bit and back to 32 bit again.
http://www.delback.co.uk/wavrep/

While the objections expressed are true, there is a program that does much as you requested.

When I tried it, I came to the conclusion that it doesn't detect clicks any more reliably than Audition does - which, in fairness, its author concedes is a general problem.

The difference is that it will detect, under various different criteria, but will not destroy your file by 'fixing' all those instances of over exuberant audio transients that are not errors. It marks them, you get to decide if they are good or bad. That is what was requested.

This won't necessarily give one superior audible results. It always depends on the music and the damage thereto, but it is often the case that many of those unnecessary fixes don't really hurt anything as far as what one hears. However, some people are finicky beyond what they can really hear.

A note to the future: the program's author claims to be re-writing the program in response to many requests to handle higher resolution files (apparently not a trivial task) although he claims that vinyl's high noise floor makes working in > 16 bit completely unnecessary. It would be nice to be able to make an objective evaluation of that viewpoint, but my experience suggests it is probably at least true that using 32 bit files buys nothing when the LP is in fairly good condition (thus requires less processing).

[sure]

It's a pity no-one has written a pattern match algorithm (à la neural nets) for clicks. I assume that Audition and other sound editors have to use the tools available, and I further assume that the tools are closely related to dark and dastardly fourier matrices, but risetime is nearly always a giveaway, whether it's longitudinal, vertical, or cross-groove damage - it always sticks out like the proverbials.

I haven't had a good look at Wave Repair, but I will, because I'd at least like to get a bird's-eye view of the likely transient profile of various parts of the recording. At the moment I do all my basic recording at very high sample rates and depths, but sooner or later I have to convert to 44.1k/16 bit. I've got great click and noise profiles that work beautifully for the bread-and-butter transients and noise at high bit rates/depths, so I'm happy to downsample earlier if I get to see what the heavy click work is likely to be like. But I still have to listen to the final product to be sure!

I hear what you're saying about sample depths for most recordings. There's nothing quite as relaxing as waiting for any declick or nr algorithm to wade through bits and bits and more bits. (I drink a lot of coffee).

I deal almost exclusively with really poor vinyl and tape recordings, so I want to get the noise in the first place and then deal with it on my terms, rassling it to the undo list, than risk not getting everything off the groove. 'Course, there comes a time when I have to compromise between speed and quality. But since I only have to please myself (if I'm happy, the customer is always happy   ), if I can find any tool that simplifies or alerts me to possible issues with a recording, I'll try it once.

Then again, I could probably just sample 11k/8 bit or stick it in MP3 format and the majority of my customers would be happy as pigs in sh!t. It's the other 25% I have to worry about, and I almost never know which category a customer falls into, unless they're wearing a hearing aid when they drop off their albums. And usually not even then  

I've only used the "fill single click now" function on occasion, so I'm mostly asking this out of ignorance.  Would splitting your file into 3 or so different frequency files (low, mid, high), and then running a click and pop check on each frequency file, improve the process at all (e.g., WaveRepair; Younglove's Decrackle Technique)?  Once each of the three files is cleaned to taste, you could mix-paste them back together.

Obviously, the split point you use to create the low, mid, and high frequency files should vary according to the nature of the material you are working with.

"Fill Single Click" can remove noises from a myriad of sources.  I consider it to be the One-Click Wonder.  Whether band splitting would help would depend on the type of intrusion.  Having used this feature extensively, I've never thought of going about it this way.  The multi-band approach would probably be most benefial cleaning LPs where a pop can have a short treble excursion but a longer lower-frequency component.  However, if the Click/Pop removal tool depends on the full frequency spectrum to analyze what is a click and what isn't, it may backfire.  But "Fill Single Click" would surely still work fine.  Someone please try this and report back!

I've actually played around with spectrum splitting and working on the various "band" files separately. From an (admittedly) subjective perspective, single-click removal didn't noticeably improve the results overall. Some clicks with a large high frequency (>12..30k) component combined with a large low frequency (<800Hz) component resulted in audibly better reduction results (less effect on the background audio, since the low frequency component (what I call the 'knuckle' ) tends to be spread over more samples than the high frequency component (what I call the 'spit' - I know there are proper terms around for these, but it just makes it easier for my limited meat CPU to give 'em colourful names   ).

The problem I found was the distortion and ringing effects introduced by the bandpass filters to begin with, and what sounded like TIM distortion when the files were remixed, plus the significant effort in ensuring that both bands' click reduction was synchronised, meant it just wasn't worth the effort.

Maybe others have got better mileage?

If there are any technical terms for band-split anomolies, I don't know what they are.  So "Knuckle" and "Spit" work quite well for me!  I like that.

There is one incredibly accurate way to band split high quality audio, which when summed back together no distortion or phase shifting remains.  Everyone should use the "Frequency Band Splitter" effect in Multitrack View, and make sure it's all coming out 32-bit.  Then fix all those clicks you want to with no sonic penalty.

I did download and try out Wave Repair! THanks for the suggestion!

It actually did find many of the clicks I told it about, and it was pretty cool jumping from the last matched click to the next and vice versa. It was also pretty cool defining the click 'style' (fast risetime, fast fall, spike, knuckle, etc), but I would have enjoyed being able to define multiple 'styles' simultaneously, with their own parameters, so it could find all the click types I needed. But there was at least some flexibility there. Unfortunately, the wave drawing routines got utterly and completely lost in all but the smallest wav files, and it was utterly impossible for me to 'see' the clicks it found without chopping the input file up so Wave Repair could draw it properly.

While Wave Repair seemed to find more actual clicks in the original audio than did AA (using roughly equivalent parameters), it seemed to have much more difficulty than AA in terms of identifying and differentiating 'mild' clicks (actual clicks, not snare taps or guitar string squiffs) in already partially fixed audio. I realise this may be entirely due to my lack of familiarity with the interface. (That's yet another reason I enjoy working with AA - I tend to appreciate visual representations of data, so great accurate visual feedback in addition to audio verification is a gift for self-taught window lickers like me).

So the Great Undetectable Transient problem still exists. (Heck, if physicists can have a GUT problem, why can't we have a GUT problem too? )

HI TO ALL  IA'M USING  RIP VINYL  A BRITISH  PROGRAM THAT IS COMPATABLE WITH COOL EDIT   AND   AUDITION 2       NO MORE CLICKS  POPS  CRACKLES   HOPE IT IS OF USE 

Murray, welcome to the forum. No need to shout here; we read you loud and clear

"No more clicks pops crackles" is a rather general  statement; people say the same about CE/AA, and we all know that that's true for 95%. The rest is dealt with manually or with other s/w. How do the results of Vinyl Rip compare with the results of CoolEdit/AA? It looks like a stand-alone program rather than a plug -in. When you record the sound in Rip Vinyl, what are the characteristics of the wav file (16 bit, 24 bit, 32 bit float).

Thanks for the info, b.t.w. I don't think I'll try it, I'm happy with my current "declick setup".

A BRITISH  PROGRAM

Does look more like a Swiss/US developer, though...

[will]

Thanks, Murray, for your comments about Rip Vinyl. It's always good to see people getting interested in protecting their vinyl, and the Rip Vinyl program does a lot of stuff that other basic recording programs (including stuff like Nero Trax Editor, Sound Recorder, and others) just don't provide at all!

But unfortunately, RV doesn't do ANY kind of transient (CLICK/POP) reduction, identification, or noise reduction. It's just a recording tool, and that's all. If you didn't hear any clicks or pops, then either a) your vinyl LPs are in perfect condition, or b) you have some sort of lowpass filter between your turntable and the computer (maybe the turntable goes into an amplifier, and the line outs go to the computer, or the turntable has some kind of RIAA/pop filter built in, like some of the turntables hitting the supermarket chains nowadays), or maybe c) you had the speakers turned off? (I'm only kidding! ).

In case other members of the forum may be interested for passing on to their friends, Rip Vinyl is actually a very nice little program, and it's really well thought-out. It does automate some of the more difficult recording tasks that the average enthusiast would find terribly annoying and difficult if they didn't have access to Audition, Sound Forge, or other high-level editor.

It has some nice controls for allowing the user to automatically split tracks "on the fly" - you can adjust the silence level, minimum silence length, and maximum track length, which RV then uses to stop one recording session and then start another. And you can do things like clicking "Shift-Stop" to manually do the same thing.

It does a couple of other appealing things that few other tools provide: In particular, the user can specify a time domain stretch/compress, so anyone with a 33 or 45 RPM turntable can plop on a 78, and the program takes care of shifting the frequencies and sample rates so that the recorded file "sounds right". For many people, that alone would make the software worth its purchase price.

RV is definitely NOT a tool for serious audio recordings at any level, but it sure is a great "mum and dad" program that probably lets folks at least get a basic copy of their vinyl and get it on to CD or in a file (probably in MP3 format, which RV supports). And that's why I'm in the game - if people don't want to pay $35 for me to provide them with a sensational-sounding and performing copy of their vinyl that sounds better than a commercial re-release, at least they might be interested in preserving the vinyl they have. It won't be anywhere near acceptable in terms of quality (noise, transient, equalisation), but it will play in their CD player or iPod, and that's better than nothing.

The problem (in case Murray or other viewers are interested) is that the program only records up to a maximum of 44.1kHz, at a maximum of 16 bits - which is perfect for a plain CD of the actual record, but it means the recorded material will never contain enough additional audio data to permit a reasonable amount of complex frequency or sound analysis.

My other concern is that Rip Vinyl intercepts and modifies the raw data stream, which is just anathema to me, although other forum members might disagree. I don't like the idea of working only with a buffer's worth of data and limited time to do whatever sample rate decimation I need to before the buffer is refilled. I would guess that's the main reason RV doesn't try to do any serious click elimination or noise reduction - it just doesn't have time (or enough data) to do that kind of analysis properly.

But at least it does let people who may not know enough about real sound editing to get interested, get involved, and who knows? Maybe some of the users will be interested enough to learn more about this work and what it means, and get themselves involved in Audition, or SOund Forge, or Diamond Cut.. or...