I Fixed the Worst Vocal on the Internet
I paid for the worst vocal I could find on the internet. Out of time, out of tune, and so dull it sounded like it was sung with a paper bag over the mic. Then I built a track around it in Suno so there was something to mix it into.
Most vocal tutorials start with a clean, well-recorded studio vocal. That is not what most people are actually working with. So this one starts with something genuinely broken, the kind of rough take or AI vocal you might be trying to rescue right now, and takes it all the way to something you would happily press play on.
Here is the full process, step by step, in Logic Pro.
Step 1: Lock the timing before you touch anything else
The backing track came out of Suno drifting out of time. I asked for 80 BPM and it did not give me exactly that, which is the classic reason AI only gets you about 80% of the way there on anything.
So the first job is timing. Using Flex Time, I pulled the instrumental so it sits perfectly on the grid at 80 BPM, then synced the vocal to that locked track.
Do this first for two reasons. Any time-based effects you add later, like delays, only sit right if the whole track is locked to the grid. And a vocal is far easier to tune and edit when the track underneath it is rock solid.
Step 2: Fix the tuning in two passes
The vocal was badly out of key, painful to even work with. So before tuning it properly, I put a pitch corrector on it just to make it listenable while I finished the timing, set to C# natural minor.
Here is the important bit. You cannot run Flex Time and Flex Pitch at the same time, so the order matters:
- Time the vocal up with the rough pitch correction on
- Bounce it in place to print the timing
- Remove the pitch correction
- Apply Flex Pitch and tune by hand
That manual pass is where you catch what the corrector missed. There was one note as the vocal came in that was sitting flat, so I lifted it up a touch and fine-tuned it until it was bang on.
Step 3: De-ess before you EQ
The vocal was dull, so the instinct is to reach straight for EQ and brighten it. Do not do that yet.
It also had harsh sibilance, those sharp “ss” and “th” sounds. If you brighten with EQ first, you make the sibilance worse. So de-ess it first, then brighten. I used Logic’s stock de-esser to take that edge off.
Step 4: Brighten and tidy with EQ
With the sibilance under control, now you can EQ. A Channel EQ to brighten the dull top end and tidy up the tone. Nothing dramatic, just cleaning it up and opening it out.
Step 5: Control the dynamics with multiband compression
The vocal was jumping around a lot in level and tone, so a single compressor was not enough. I split it into bands so each area could be handled differently.
The key move here was pulling the muddy mids down. The vocal was very muddy, so reducing that region cleans it up and, as a knock-on effect, lets the top end breathe. That means you do not have to keep cranking EQ to get brightness. You control the problem instead of masking it.
Step 6: Glue it back together
After all that multiband work, the vocal needs pulling back together into one cohesive sound. A standard compressor does that gluing job. I used the Studio FET, set by ear, to hold everything in place.
Step 7: Add presence
Almost there. To bring the vocal forward and give it that finished, present quality, I added a Vintage Console EQ and lifted the presence as I listened.
Step 8: The reverb bus trick that keeps vocals clean
This is the move that makes the biggest difference, and it is the one most people get wrong.
Instead of putting reverb directly on the vocal, set up a bus. A bus is just a separate channel where you load up effects, then send a portion of your dry vocal to it. That way you control exactly how much wet signal you get.
On the reverb bus I chained an EQ, then a reverb (the Quantec Room Simulator on a medium room setting), then a compressor. I only sent about -15 from the dry vocal, and I shaped that send so only the middle of the vocal feeds the reverb:
- Low cut at 200Hz
- High cut at 5k
That stops the reverb muddying up the low end or adding hiss up top. Only the body of the vocal gets the reverb.
Now the clever part. A vocal drowning in reverb all the way through sounds washed out. So I want lots of reverb, but only when the vocal is not singing. To do that, the compressor on the reverb bus is sidechained to the lead vocal.
The reverb bus now listens to the vocal. When the vocal comes in, the compressor clamps the reverb down. When the vocal stops, the reverb blooms back up into the gaps. The settings:
- FET compressor, ratio around 4:1
- Fast attack to grab the transient instantly
- Fast release so it lets go quickly
- No makeup gain, auto gain off
- Threshold set to pull down about 10dB
The point is not to kill the reverb. It only ducks by around 5 to 6dB. Just enough to keep the vocal clean and present while still getting a big, vibey space around it.
Step 9: The delay bus
The delay uses exactly the same approach. Another bus, with an EQ, then a stereo delay. The preset I used is the streaky vox delay, which is one of the things included in the session file you can grab above.
Then I copied the same sidechained compressor straight off the reverb bus onto the delay bus. Identical settings. The only difference between the two is one has reverb and one has delay. Same ducking trick, so the delay stays out of the way of the vocal and fills the spaces around it.
The result
That is the full chain. Timing locked, tuning fixed in two passes, sibilance tamed, tone brightened, dynamics controlled with multiband then glued back together, presence added, and a reverb and delay bus that duck out of the way of the vocal instead of burying it.
A vocal that started as the worst one I could find on the internet, sitting in an AI backing track that would not even stay in time, now sounds like something you would press play on.
