This is a quick showing of a personal project. This post is not meant to be an in-depth article or guide -- that will come later.
Below is an audio sample from an old nth generation VHS (a copy of a copy of a copy), converted to DVD with a lousy early E-series Panasonic DVD recorder in 6-hour mode. Yuck! The video was nearly ruined -- but that's another topic for other posts/guides. (And I modded a custom Avisynth filter in the process.) This is just the audio portion.
Listen now to the attached before clip. Turn it up nice and loud to hear all the noise.
Terrible, yes? It has extreme hiss and a mild buzz that ruins the audio. The noise is about half as loud as the audio. So 33% of the audio signal is just noise, and that needs to be removed. The overall volume is also very low to begin with.
How to Restore the Audio
Understand that restoring video/audio is rarely as easy as using one program and pressing a "restore" button. It doesn't work that way. That's only possible on the CSI TV show (fake!), and is amateur wishful thinking.
For a project like this, the initial plan is often:
- 1. isolate a noise print with
Sound Forge
- 2. remove the "bad audio" (noise) with Audacity, while trying to maintain the "good" audio
- 3. correct remnant noise and/or Audacity-added distortion in
Sound Forge
More steps (or different steps) are possible, if needed, but this often does the trick
Step 1 - Noise Print from Sound Forge
Finding a viable noise print is one of the more tedious aspects of this restoration workflow. It needs to be a "silent" clip (meaning the "good" audio), and have only noise. Sometimes that's not possible, so another (often harder!) workflow must be used.
This project had many files from the same source. So the noise was identical on all clips. Thought JBB1 did not have a good print, one of the others did.
Listen to the attached noise print.
Noise prints need not be longer than a few seconds. Even a print that lasts for less than a second can often be good enough. The longer the print, the more accurate, of course. But we don't always have that luxury. You use what you can. For example, the JBB12 clip had two excellent sections of only noise.
jbb-noise-print.jpg
Just copy the print/clip to a new file:
- Highlight the noise-only section, CTRL+C to copy.
- Create a new file (of the same specs), CTRL+V to paste, save as, and now we're ready to restore some audio!
Step 2 - Pass 1 in Audacity
The Audacity step is both helpful and destructive. While it removes noise, rarely is it flawless. This step is mostly easy, yes, but you're still not left with a usable file. It'll need more work.
File > Open the noise print, and again File > Open the audio file to be restored. DO NOT DRAG AND DROP! Each instance should open in its own window. If both open in the same window, you did it wrong.
Removing noise is an easy process -- but with caveats. There's only 4 settings, but those can make a big difference with how the audio sounds. The defaults are often not best, and (as always, in the world of video/audio restoration) you'll rarely use the same settings twice. Too many novices just use the defaults, which can really muffle the leftover audio.
CTRL+A (select all) the noise print, then Effect > Noise Removal. Step 1 engages the noise print in memory.
Now go to the main file (not the noise print), CTRL+A the audio, go to Effect > Noise Removal, and preview it. If the preview is not helpful, cancel, and select a good audio portion to test listen. Effect > NR, now preview. The more through guide comes later, but as an example, this was the setting used for the attached JBB1 re1 (restore pass 1) file:
| You must be logged in to view this content; either login or register for the forum. The attached screen shots, before/after images, photos and graphics are created/posted for the benefit of site members. And you are invited to join our digital media community. |
None of these settings is aggressive, which is what you want. The lower, the better, while of course removing enough noise to make it useful. Don't try to remove 100% of the noise, but anywhere in the 90%'s is good.
What we're left with is a file with noise removed, but with distortions where that noise was removed. It's still not usable, and will cause speakers to crackle. The better the speaker (reference, studio grade, etc) the worse it sounds. So never assume it's "fine" if you're using cheapo speakers -- for example, built into a laptop, or $10 from Walmart.
Step 3 - Pass 2 in Sound Forge
By far the most difficult step is making decisions in Sound Forge. Right away, you'll need to decide which type of filtering corrects the noise -- the Paragraphic EQ, the Audio Restoration filters, or (less commonly) the Graphic EQ. Each operates differently.
We've provided many Sound Forge filters for free download here:
http://www.digitalfaq.com/forum/vide...io-filter.html
... but it's not uncommon to need new ones. Every month, we tend to make new filters for projects. Again, video/audio is not a one-size-fits-all type of project. Sometimes past filters work fine, but sometimes the noise is so unique that it can only be corrected by a new custom filter or preset/setting. And this hobby project was no different:
jbb-nr-step1.jpg
| You must be logged in to view this content; either login or register for the forum. The attached screen shots, before/after images, photos and graphics are created/posted for the benefit of site members. And you are invited to join our digital media community. |
Two filters were created, and the exact distortion level of the Audacity output determines which is used per file. On the first clip, the second Paragrahic EQ preset was used.
Listen to the attached re2 file.
This file had very weak audio values that needed to be normalized. After testing the normalizing both before and after the Restoration filter (after was best), the file was processed in only two steps. And for Sound Forge work, that's unusual! More steps are often required!
What I'm left with a balanced audio file -- almost no noise, decent tremble/bass/midtone values, and 50% volume saturation (normalized). It's something that can now be heard and enjoyed, not simply heard and tolerated.
Again, this post is not meant to be an in-depth guide -- just a quick peek at a personal project of mine.
Sound Forge is professional software for a reason. It's not dummy-friendly. It has a learning curve. Not only do you need to learn the software, but you need to understand the basics of audio to truly make it a useful tool.
This is an example of the type of results that are possible.
Conclusion
Again, this is just one possible workflow for restoring audio. For a few years, it's been our primary restore method, useful for about 50% of all audio restoration work. That may change someday, if/when better software exists.
That other 50% of audio restoration, which cannot use this method, is splintered into at least a dozen different possible workflows. It can range from Sound Forge-only work (with complex filter chains!), to workflows that require specialty software -- even workflows that require the same program to be used more than once! These are the truly challenging projects, though I find the challenge fun.
I've been known to use the more complex methods on audio samples posted by our Premium Members.
Important Note
This is just my hobby project. The main reason I write guides is to help fellow hobbyists -- collectors of old TV shows, cartoons, movies, concerts, documentaries, etc.
But when it comes to family memories ("home movies"), most people do a huge disservice to their family and loved ones by using low-end hardware and novice methods. Something that important really should be professionally transferred using the best hardware and methods that exist. When the project is important, don't DIY (do it yourself), let professionals handle it. It's not easy work.
Consider letting the professionals at
The Digital FAQ restore your videos, and handle your cherished family memories.
