Simon Billington
Forum Replies Created
-
Not a bad effort actually, Steve. It does still tend to distort a bit especially on the right side and is a little resonant. On the whole though it is an improvement on the original.
I guess it might help to mention I also ran the source through a Dereverb algorithm set at around 3 secs before trying to enhance it. In retrospect as i listen to the results again, I do notice the enhancement brought up the noise a bit more than I initially would like. So if I were to do it again, I’d attempt to look for a happy medium between the repaired version and the enhanced version.
Another way of handling that is running the repaired and the enhanced version side by side, as long as they are phase coherent, and blend a balance of the two. I do that often, not just with restoration, but even things like color grading, I might suddenly decide its too much and blend the grade back with the original. I like mix knobs on compressors for the same reason.
The actual settings I used might be hard to trace back since I did multiple processes, many times with the same module, and I didn’t save any of the settings on the way. Often I do, but in this case I didn’t, since it was more of a curiosity.
In enhancing it I do also remember doing some subtractive eq on it, as the first stage of enhancing. I rolled the bottom off at 35Hz, and pulled out 2.5-4.0dB at around , 85, 170, 330, & 650Hz . I used a little of Noveltech’s Vocal Enhancer, probably a bit more than I should have. The idea is to add intelligibility and boost overall presence in a more transparent way to eq boosts. This is how Noveltech describes their plugin…
“Based on Noveltech’s patent-pending Intelligent Adaptive Filtering (IAF) technology, VOCAL ENHANCER works on both frequency response and dynamic properties. It identifies and enhances the characteristics in source audio material that are pleasing to the human ear, and gives users control over the perceived definition and depth of your vocal track’s characteristics”
After which I applied a bit of compression with H-Comp to bring the dynamics under control, but not too much as in to take away from the pulse of the beat. The Punch Knob was actually handy at allowing be to blend a little bit of the snap back in. Then it was tiny bits of additive eq, nothing more than 1-2db and some multiband enhancement from Vitamin to add a bit more weight. I didn’t know I was going to use it at the time, but If I had, I would have used a little less of the Vocal Enhancer.
I had another shot at trying to remove more distortion that process seemed to have brought out a bit and I even rotated the phase of the audio just to get that tiny bit more of definition out of it.
Anyhow, the way I seem to find works best in most things is not processing hard with a small handful of processes, but using a subtle blend of more, especially in difficult cases. That way the changes appear to be much more transparent. Notice how there is much less resonance in the final result, less boom, its not harsh on the ears, yet its much more clear and intelligible. It did bring the noise back a bit too much, so I would ultimately go for a blend of the restored and enhanced process.
I feel its important also the process doesn’t rob the source from too much of its “vibe”. Something I feel important to preserve especially when handling memory vids like weddings. People want to be taken back every time they watch it, so they want that feeling of actually being there again. My job is just to enhance that experience so they can romanticise about it.
Anyway, that’s my take on it.
-
No, I had to crack out the big guns.
I did it all in RX Advanced, but with the enhancement I used 3rd Party plugins, mostly Waves.. Mainly H-EQ, H-Comp & Vitamin, that I can recall.
-
Thats a very difficult issue to solve indeed.
I downloaded the sample to have a look at it and see if I can understand what’s going on, but much of it still remains a mystery. I’ve never heard audio issues quite like this before, but I decided to have a crack at it.
I ran the original distorted audio through a few restoration processes. The end result isn’t perfect, it rarely ever is, but it’s certainly much better than the original. It should certainly suffice for the wedding vid.
The caveat is that it did take me a little time to nut out and it wasn’t exactly a straight forward process. In the end it involved a combination of hum removal, declipping, and multiband declicking. I’m not even sure Audition is even capable of half of this. I had to go hardcore with something much more expensive.
In addition to having the noise removed, this file has undergone an enhancement process to improve intelligibility and overall quality of the audio. You can check it out here.
Unfortunately, the enhancement process brings a few more of the imperfections to light, but barely. I still think its a more preferable trade off over both the original and the restored version above and should increase the enjoyability for the viewer.
-
It would be a shame to remove all of the volume automation just to remove the noise out of the audio, but this is getting a bit tricky.
You could edit the source files and the will automatically update in the project, except it gets risky. If you over do it you would want to have a backup. That in itself isn’t hard. If, however, you had to restore your originals from your backup you would have to replace all the clips manually. That’s the problem.
What you can do is duplicate your project, but give it its own dialog assets, so its not linking to the same files as the original, then remove the noise on those source files.
If you’re dialog is still tied to the clips, just unlink them, even convert them to regular audio files if you have to.
Adversely, you could always use a 3rd party plugin on the audio. Something like iZotope RX Dialog Noise Remover, or Waves NS1/W43. That might be good enough to get the job done without any issues.
Just be aware those plugins amy induce latency, so make sure the audio remains in sync at the end of the day.
-
Traditionally its done with a Low Pass Filter, yell often find one on an eq.
If you engage it, the more you lower the frequency, the more you start losing the high end or those treble bits. If you start to lose intelligibility raise the frequency again.
If you felt like spending some cash, this is an easy thing to use that does the same job. Just make sure you set the Resonance to “None”.
I’m all for real world physical experiments, but this is somewhat of a quick and easy thing to pull off with an eq, unless you want to try going into it in more detail. But then time does become an issue.
-
Ah yeas. Well that was my going theory, but it seems you got it sorted.
-
Okay. It’s not perfect, these kind of repairs rarely are.
The audio glitch is almost completely gone. Its certainly good enough for mainstream.
10313_implementationedit.wav.zip
It is however missing 0.24ms from the duration of the clip. It was really the only way it could be fixed. Hopefully, its not an issue. if it is there is something we can do about that too.
-
All though there is this plugin here, which is more expensive and more complicated, but will offer all sorts of creative options for sound effects if you ever need it.
https://www.waves.com/plugins/metafilter#
I took the liberty of reproducing the setting that you would need for 3 sec drop outs. However i think you would have to adjust the Rate for what works with you, the one above the Sequencer. I think it’s defined by the Tempo of the project. I had to use around 4.0 in this case. I thought this was a bit odd and I certainly don’t think the number represents seconds, more like beats. You can adjust it to what suits you.
If you do go this way, make sure you pull the Output down to stop it from clipping. The way this is set up is it boosts the signal as well to create enough difference between loud and quiet.
The Sequencer is what controls the actual drop out, so if you want a longer period of silence, pull down another fader. If you want a shorter period, pull one of them back up to the top.
-
Although Color Finale has LUT features now.
Wish I could say the same about Red Giant Colourista III. It has its owns strengths however. Who knows, the competitive nature of color grading plugins may force them into including a LUT facility in Colourist the next time around.
-
If you want to do something that long and insert a silence in every 3 secs you’re going to not do it by hand, that’s for sure. You do need an automated process, thats for sure and I’m fairly certain thats what you were hinting at.
Unfortunately there is no way I know to do this in Audition. Logic however has a “Tremolo” plugin which you should be able to utilise to such an effect. Given this is Audition you can get yourself a cheap Tremolo VST plugin that allows you to achieve such an effect.
You need to be able to select a a time in seconds for it to cycle, you will need an asymmetrical looking square wave similar to the one in the main pane of the plugin in this pic. In Logic this was achieved by setting the Smoothing parameter to 0 and Symmetry up to around 80 or so.
The Depth and Mix (or Wet/Dry) if it has one you will want to set all the way up. Phase should be 0, and the Rate should either be 3 seconds or 0.33 Hz, which works out to be a 3 second period.
If you use the Tremolo in this manor it will drop audio every 3 secs for a short period. You then just need to render the VST and it will effect the entire duration of the audio.
