Audio Blisters

Check your audio levels, then check them again

I loved this week’s CodeNewbie Podcast episode. We had Stella Cotton on the show to talk about site availability. She nailed it. She got technical, personal, vulnerable — the three angles I try to hit as an interviewer, and she made my job super easy. She offered us her knowledge, her stories, her experiences with a deep generosity. It was humbling. It was beautiful.

Unfortunately, the audio quality was terrible. I blame us, my husband and I, the producers of the show. Anything that goes wrong technically is 100% on us, and we failed this interview.

So much noise

Just about every mic comes with noise. Record yourself on your laptop, play it back, and you’ll hear a soft, soothing “shhhhhh”. Most of the time, it’s not a big deal. If you have a fancy mic with a mixer and pre-amp, you can avoid having to deal with this noise at all. But for the rest of us, we get rid of it in post-production.

We use Adobe Audition to edit our audio, and we use a process called “Noise Reduction” to deal with the noise. But since you’re speaking over that noise, removing the noise means that your voice gets processed too. If the noise is really low, it’s not a big deal, and you can barely tell that your voice changed. Occasionally, you’ll get artifacts (if you listen really hard after doing noise reduction, you can hear the breath you take before you talk sound every so slightly robotic), but for the most part your voice is fine.

Unfortunately for us, this lovely, soothing “shhhhh” that we’ve always worked with was really loud. Really, really loud. So loud, it sounded like Stella was standing in the rain. We had to process her voice a ton, and you it’s very obvious. If I didn’t tell you, you’d assume a bad internet connection or a bad mic, and in a way, the source of the problem is the mic. But the robotic, metallic filter on her voice is the processing we applied.

Clipping

At some point in the interview, the mic lost its mind. Her voice came through robotic and distorted, and I couldn’t understand her. This has happened a few times before, and we ask the guest to unplug the mic and plug it back in. Unfortunately, when the mic was reset in this way, it also reset her audio levels, and we didn’t check her levels again.

Her levels were too hot, which means that her voice came in so loud that it clipped, creating a terrible, harsh, distorted sound. Clipping is the nails-on-the-chalkboard of audio. It literally gave me chills. Her levels were so hot that I caught 40+ moments of clipping that needed to be processed.

You can’t truly fix clipping. If it’s very slight, you can use something in Adobe Audition called the De-Clipper, which takes the shrieky bite out of the sound to make it barely bearable. But it can only do so much. Most of the processing I did resulted in converting the shriekiness to a muzzeled “zzzhhhh”, which was just as bad but less painful.

And here’s the thing that really sucks about using the De-Clipper: it takes up a ton of space. I started with 3GB, and ran out of space. I moved things around then added 6GB. Then 10GB. Then 26GB. At 26GB, I barely made it without running out of space, scrapping by with a gig to spare. It’s also incredibly slow. The time required to process just a few seconds of audio ranged from 1 to 3 minutes. With 40+ instances of clipping, I spent about 2 hours of my editing time just waiting.

Cutting Her Laugh

I did the best I could, and it sounded terrible. The artifacts and the clipping were a clear distraction from the awesome things that Stella was saying. So I made the difficult decision to edit out the parts that clipped. I don’t like editing for content. I like to preserve the interview as much as possible. The purpose of our editing is to removed the obvious mistakes and dead air so you can focus on the interview itself. But the clipping was too much. I removed words, stitching together broken sentences, maintaining the integrity of her words as best as I could. But I didn’t like it.

One moment stands out the most. She laughs between two thoughts. The laugh was too loud. It pierced and screeched, and it had to be cut. I took the two ends of her thoughts and glued them together, leaving her words in tact. But without that laugh, it was a different sentence. The laugh changed everything. The laugh introduced a playfullness to the point, a note of introspection. It broke her thought and forced you to deconstruct her words along with her. Without that laugh, the sentence was too serious, too purposeful. It was everything she said, but nothing she meant. My heart broke cutting out that laugh. I knew that dealing with clipping was tough. I knew that it’s one of the worst audio mistakes you can make, apart from forgetting to press record. But I didn’t think it would put me in an editorial bind.

I spent the entire night fighting sounds and decisions. Eight hours of editing later, I produced a 52-minute episode full of incredible content, spotted with audio blisters and bandaids. I opened my bottle of reminders and popped this one in: a note to check my levels, to check them twice, and check them again. I was exhausted and disappointed when I finished, hoping her powerful story would reach people before the terrible quality left a mark. I cringed as I hit publish at 7:43AM, heart heavy and lesson learned.