A retired singer and producer wanted his own vocal to sound like the one on a Suno AI cover of his song, and to my ear the difference came down to heavy reverb and parallel compression.
A client brought me an interesting question for the age of AI music. He’s a retired singer and producer who has written hits over a long career, and lately he’d been putting songs he recorded 20 or 30 years ago into Suno to generate covers in a modern production style. On one of them he preferred Suno’s mix to his own, especially the effects on the vocal. His plan was to have Suno make a backing-track version of the song, sing his own vocal over it, and get his voice sounding the way the AI singer’s did. He didn’t know how to get there with Logic Pro and the plugins on his system.
In his studio days other people set up the vocal chain for him, and he gave them the direction or the OK. It’s how studios divided the work in those days. This time he’d heard exactly the sound he wanted, but he had no route to it.
Hearing the difference
A lot of my work is plugin installs and configuration problems, so this one was interesting because I had to use my ears. I could have gone to Logic’s track presets, picked something like a male rock vocal or a big vocal hall, and skipped through until one sounded close. It’s one way of doing it, but I don’t enjoy skipping through presets. Instead I listened to the two recordings side by side and picked out the sounds that were different.
To my ear, the Suno vocal was swamped in reverb and had parallel compression on it. It sounded bright, but still kept its dynamic range.
Building the chain in Logic
I set this up in his session and showed him how it works:
- Two reverbs: a shorter one for a room-type sound, and a very large one for a long tail.
- A compressor on the vocal track itself, levelling it off slightly.
- A bus send from the vocal to a second compressor set to compress very heavily. I mixed that squashed copy back in under the original, along with the two reverbs, for a thicker sound.
- An EQ boost for warmth. Flicking between the Suno vocal and his, I could hear his was lacking it.
Both vocals had been recorded on good microphones by a good singer, even if the AI one was virtual. With the changes his own vocal was closer to the Suno version, and he was delighted.
What interested me is that the AI had made something he preferred to his own version, and because it’s a black box he had no way of seeing how it got there. Getting there took someone who could hear what had been done. Taking a mix apart by ear like this feels like a dying art. I’m far from a world-class sound engineer, more a jack of all trades with studio technology, but I’ve been in enough mixing sessions to know what I’m hearing.
When an AI track has a vocal sound you’d like on your own recording, it’s worth listening for the reverb, compression and EQ on it and rebuilding them with the plugins you already have in Logic.
Not sure it’s the same problem? Ask a question first
