Sometimes the Best Restoration Tool Isn't a Restoration Tool: Fixing Audio Artifacts Without iZotope RX

One of the biggest misconceptions about audio restoration is that every problem has a corresponding plug-in, module, or  button.

Need to remove noise? Use a noise reduction plug-in..

Need to remove a click? Use iZotope Rx de-click. 

Need to repair an artifact? Open your favorite spectral editor (have a pick of many!). 

And to be clear: modern restoration tools are incredible. I use them regularly, and they can save hours of work.

But every once in a while, a problem appears that reminds me of an important lesson: Sometimes the best restoration tool isn't a restoration tool at all.

Recently, a client sent me a clip from an archival documentary and asked if I could help repair a strange artifact. They had already tried using iZotope RX, but weren't able to get a satisfactory result.

The first thing I did wasn't open any software. I did what I always do first: I checked the source.

Whenever I encounter an artifact, I want to know whether it was introduced during transfer, conversion, downloading, or editing. If the problem exists only in my copy, the solution may be as simple as finding a cleaner source. This has saved me many times. 

But in this case, I located a second copy of the documentary and compared the two. The artifact was present in both.

That told me something important: the problem likely existed long before the file reached me.

Now the real work could begin.

Listening closely, I noticed the artifact occurred during the word "ounce."

Rather than focusing on the artifact itself, I started thinking about the word.

What was being said? How was it being pronounced? Was there another place in the recording where a similar sound existed?

A few moments later, I found another word: "violence."

Different word. Similar ending. Both contain a very similar "nce" sound. And suddenly, the problem stopped looking like restoration and started looking like editing.

The idea was simple: What if I borrowed the clean ending from "violence" and used it to replace the damaged ending in "ounce"?

Unfortunately, there was another complication: music was playing underneath the narration.

If I copied the replacement directly from the original recording, I'd be copying the music as well. Any mismatch in timing, rhythm, or ambience could make the edit far more noticeable than the artifact itself.

So instead of immediately editing, I separated the narration from the music using a stem splitter. I happened to use SpectraLayers, but there are many tools capable of this today, including several affordable and even free options.

Once separated, the solution became straightforward. I exported the stems, brought them back into Adobe Audition, copied the clean "nce" sound from "violence," and carefully edited it into place. A little fading. A little envelope shaping.

The result was remarkably transparent.

What looked like a complicated restoration problem ended up being solved with a fairly traditional editing technique.

And that's the lesson: tools matter - but knowledge matters too. The more you understand about language, editing, timing, masking, and problem-solving, the more options you have available to you.

Sometimes the newest AI-assisted restoration tool is the right tool. But knowledge and creativity are what turn tools into solutions.

If you'd like to see the process in action, I've included the video walkthrough below.