Accusonus Uses AI to Make Quick Work of Bad Audio
Accusonus is making audio repair easy for sound and video people using machine learning.
I do not consider myself a sound editor or an audio expert, but as a picture editor, I understand the importance of good sound throughout the edit. Bad sound makes a good picture edit look less good, so cleaning up your audio in the offline edit is extremely important.
Hopefully, you record great sound for your videos and podcasts or are working with people who understand the importance of great sound and will prioritize it on set. That is what I wish for all film, video, and podcast-makers, but alas we are often given poorly recorded sound to work with. That's where Accusonus can help.
As one of the first audio engineering companies to use machine learning and AI technology in an audio repair plugin, Accusonus makes it to where you never have to leave your video or audio editing software to use it. The ERA 4 Bundles (Enhancement and Repair of Audio) are packages of single-knob audio repair and leveling plugins with multi-platform compatibility that helps picture editors, sound editors, and podcasters work a whole lot faster.
I tried these plugins within Avid Media Composer, Premiere Pro, and Davinci Resolve and was impressed with how good I could get my audio to sound without having to do very much at all. In fact, it's the simple turn of a single knob that does the trick in most cases for poorly recorded audio.
Let's go over each plugin.
The Noise Remover plugin is the tool I find myself using most in my video edits. This tool is exceptional at removing constant sounds: air conditioners, fans, generators, refrigerators, etc. – but I even tried it out on exterior New York City street scene and a dialogue scene on Rockaway Beach. The plugin was able to clean up those tough outdoor scenes, exposing the dialogue from underneath the constant wind and ocean, without my having to change any parameters. I found 50%-75% to be the most applicable range.
The Reverb Remover corrects the effect of bad acoustics for on-location sound, eliminating the hollowness of a room and bringing the speaker’s voice closer to the mic. Think: a boom microphone recording a hallway dialogue scene. This plugin removes the room from the audio and makes a boom sound more like a lavalier.
The De-Esser does exactly what it sounds like: it removes emphasized "S" sounds from words in dialogue or interview segments. What makes this plugin different from other de-essers is that you can actually see the software detecting the S’s in your waveform.
De-Esser highlights the affected parts of your waveform by targeting the frequency range of the S’s, and it will only reduce that frequency. The plugin even takes into account the fact that women’s voices often have S’s on a different frequency, and you can change the frequency you’re targeting by clicking ‘Normal’ or ‘Broad.’
The De-Clipper detects and corrects clipping in your audio for that over-enthusiastic performer who speaks too loudly into their microphone.
The Plosive Remover removes pops and spikes in your audio caused by P’s, M’s, and B’s. You can see the plosive being targeted by the highlights in your waveform. Normally, I would go into my audio clip and draw keyframes around the plosive to manually lower the levels around that particular peak in the waveform, which can be quite tedious. This plugin saves a lot of time.
And on the subject of major time savers, the Voice Leveler automates a typically labor-intensive process of normalizing different voices in dialogue. Think: a three-person dialogue scene but only two lavaliers, so two voices close to the mic and the third voice sounds far away. To further test this, I spoke into a microphone and then walked five feet away and continued speaking. The plugin was able to level out my voice to sound like I was standing in the same spot the entire time. Another interesting feature of the Voice Leveler is that its Breath Control can detect breaths and not include them in the leveling.
You can see in the waveform that the plugin detects smaller waveform and raises it to match your levels. Using the plugin on a track-level, rather than clip-level, is especially helpful if dialogue is organized on its own track. As a very visual person, I like that I can actually see the plugin working. In the image below, the black waveform is my input – the recorded sound, reflecting peaks in waveform when I was standing next to the microphone, and then almost flat waveform in the center when I walked away. The yellow is the output, the repaired audio reflecting normalized levels.
The Accusonus ERA Bundle is fully supported by the following audio and video editors:
- Avid Pro Tools 12.6 (or higher)
- Audacity 2.2.2
- Apple Logic Pro 10.4.3 (or higher)
- Ableton Live 9 (or higher)
- Cockos Reaper v5.9
- Image Line FL Studio 12
- Presonus Studio One 3 (or higher)
- Steinberg Cubase 8 (or higher)
- Adobe Audition CC 2017 (or higher)
- Apple GarageBand 10.3.2
- Adobe Premiere CC 2017 (or higher)
- Apple Final Cut Pro X 10.4 (or higher)
- Blackmagic Design DaVinci Resolve 14 (or higher)
- Avid Media Composer 2018.12
- MAGIX Vegas Pro 15 (or higher)
Price and Availability
These tools from Accusonus are available now. Here's a pricing breakdown of each bundle:
[Editor's Note: From now until December 1st, you can get both the ERA 4 Standard and Pro Bundles for up to 73% off during Accusonus' Black Friday Sale. Get the ERA 4 Standard Bundle for just $99 and the ERA 4 Pro Bundle for $299.]
To learn more about Accusonus and the ERA 4 Bundles, check out their website.