Accusonus develops audio software tools that are based on machine learning algorithms. The company has released the ERA Voice Leveler in October of last year. That was a single-knob sound leveller using new patent-pending algorithms to automatically detect distracting level differences in voice recordings and correct them in real-time.
Tomorrow, the company will be releasing their updated ERA Bundle 3, the collection of single knob plugins that are said to be ideal for videographers thanks to their speed, ease-of-use and effectiveness. The updated version of the Bundle includes a brand new tool, the ERA DeClipper, an entirely automatic DeClipping software.
When I looked at the ERA Voice Leveler, it was (and is) supposed to let you adjust gain automatically by turning one knob – without leaving your DAW or NLE environment – with the AI behind it touted as capable of fixing vocal and dialogue volume problems including moving speakers, recording two or more speakers with a single microphone (e.g. interviews, lectures, etc), sudden gain changes due to audio/video edits and re-arrangements in post-production and time-varying voice-overs.
I guess the ERA DeClipper is something similar in that it will let you control the “fix” with the turn of a dial. How good the ERA DeClipper will ultimately be is something this journalist isn’t going to be trying out.
After the test that I ran with the previous “unique” module by Accusonus, I am not interested in a second try. I’ll stick with iZotope RX 7 Advanced, an app that applies iZotope’s machine-learning algorithm to a whole range of problems in the form of a Repair Assistant and of which tests revealed that RX uses three or more of its dedicated dialogue modules to fix bad voice recordings. And failing in some circumstances even then.
In short, I don’t believe in automatic sound fixing with AI – not just yet, anyway. The results are flaky and the one-button approach, however enticing that may be from the point of view of simplicity, just isn’t going to cut it.