Reminder: Auphonic is still a terrific audio tool!

It now seems half-a-lifetime ago that I used to commit an hour or two every Friday to join Randy Resnick’s VoIP Users Conference (VUC) calls. That long-running project developed a global community of like-minded geeks, addressing a broad range of technology topics.

If you’re not familiar, VUC was a live call-in podcast that ran from 2005 to 2018. There were around 800 episodes, with numerous guests from the far corners to the communications ecosystem.

It started on a service called Talkshoe, accessible via telephone and SIP. After interviewing ZipDX founder David Frankel (VUC381) about his (then) new project it transitioned to the ZipDX wideband conference platform. Sponsor Voxbone (now part of Bandwidth) was able to provide dial-in access numbers in dozens of countries. When Google introduced Hangouts, VUC naturally transitioned to that new video platform, but always maintained integration with the telephony realm.

One of the things the project required was post-processing of the recording each week, to ensure consistent sound level for all the various participants. In a podcast, dramatic changes in level from speaker-to-speaker is very jarring to the listener.

Levelator

For quite some time, we used an open source program called The Levelator to do this audio post-production. This program was written by the team at The Conversations Network, who were very early in the podcasting game.

The Levelator worked reasonably well. Running locally, it was very simple. With no user interface, just drop an uncompressed WAV file on the application window and it would work its way through, creating an optimized version.

Alas, when The Conversations Network folded its tent in 2012, The Levelator became abandonware. Although, a 2010 release remains available online.

Enter Auphonic

Of course, podcasting caught on in a big way. Eventually, others stepped up to create new tools for batch mode audio post production. One of those tools was in the form Auphonic.

Georg Holzmann, the founder of Auphonic, appeared in a VUC episode #514 on July 11, 2014 to explain what was then both an application and a web service.

I suspect that VUC founder Randy Resnick made use of the service to process the VUC recordings, such that the levels of the various participants matched as much as possible.

Time passed, and since my routine no longer included such audio processing requirements, Auphonic passed from my mind.

Positive Physics

In the summer of 2023 I came to be helping Jack Replinger, the co-founder of an online teaching tool known as Positive Physics. Jack had recorded a large number of screencasts in support of his physics and chemistry courses. Unfortunately, in about 48 of the screencasts the audio was recorded extremely low in level. So low that someone watching the lesson had difficulty understanding what was being said, even if they turned the volume on the computer to maximum.

He faced two options; try to fix the existing screencasts or spend many weeks recreating them. He asked me to explore the faulty videos. The hope was that if the audio was clean, but just very low, perhaps there might be a way to remedy the problem, making them usable. That would save weeks of work.

As is my habit, I started in Adobe Audition. Here I learned the news was mostly good. Simply normalizing the audio brought it up to ideal level.

Happily, Jack had recorded in a reasonably quiet room. Even so, such a dramatic change in gain brought with it a much higher noise floor. Disturbingly so. I could use noise reduction tools to moderate background noise, especially between phrases.

I began to process the faulty videos. The process being manual, it was time consuming. Also, in my hands, a bit variable.

This is when I remembered Auphonic! Revisiting their site I discovered a service that had been continually improved since I first saw it. It has a simple user interface, offering batch mode processing of audio and video files.

All I had to do was craft a preset that would be used to process each of the files. The preset included configuration for loudness normalization, voice EQ and noise reduction that would eliminate constant noise, consistent with the background noise created by the dramatic change in level.

While Auphonic will act on video files, in this case, I found it better to strip the audio away from the video. Let Auphonic act on a batch of MP3 files. Then replace the audio in the existing video files. Finally, upload everything to YouTube for deployment in Positive Physics.

Auphonic allows a free account to use up to 2 hours of processing time per month. For larger needs, they charge on the basis of the amount of time used to run your processing. I paid $12.80 to get 5 hours. That was enough time to do the whole project, over the course of a few days.

Compliance Announcements

Since then I’ve incorporated Auphonic into my ZipDX routine. Some of our customers require compliance announcements at the start of their conference calls. These are recordings that are played to each participant before they join the conference. They typically stipulate the specific terms the participant agrees to by joining the call.

ZipDX has a built-in text-to-speech function that lets customers make a compliance announcement using a synthetic voice to read the script. Even so, some clients are in the habit of sending me a recording made in their office. When this happens, I run the recording through Auphonic. It delivers perfect level and cleans up any background noise. So easy. It takes just a few seconds. Much easier (and way more consistent) than me spending time in Audition.

Here are before and after images of an example. The original recording is extremely low, with considerable ambience. It’s as if they recorded it by placing a smart phone on a desk, but a bit distance from a computer speaker.

Auphonic cleans it right up, making levels perfect and gating out the ambient noise between phrases. It does this in seconds. Much faster than I could do it in Audition.

Alas, since I do not hold the rights to the recording, I cannot offer them up for your listening pleasure. You’re not missing anything. They’re deadly dull. You’ll just have to take my word that they illustrate just how brilliantly Auphonic cleans up otherwise marginal recordings.

Like Minds

On the other hand, maybe don’t take my word for it. While this post sat in draft form, awaiting some inspiration to bring it to completion, I stumbled across someone else who apparently finds Auphonic to be useful.

That someone is Ellis Rovin, audio producer for MKBHD. He posted a video asking, “Can AI replace our Audio Producer” in which he describes the many virtues of Auphonic in greater depth than I can manage here.