Soothe 3 Is the Resonance Suppressor Everyone Calls the Best. I Built the One That Beats It.
THE_SMOOTHER 2 against Soothe 3 by oeksound, measured side by side on 12 things a producer cares about: THE_SMOOTHER 2 1.5.1 is ahead on 6, level on 5 and behind on 1 against Soothe 3 1.0.5. How I built it with Soothe 3 as the bar, the method, every number, and the one result that goes against me.
Ask producers which resonance suppressor is the best and most of them name the same one: Soothe 3 by oeksound, $259. It earned that reputation.
So when I sat down to rebuild THE_SMOOTHER 2 from the ground up, I didn't set myself a vague goal like "better". I picked the one everyone calls the best and made it the bar:
Nothing ships until it beats Soothe 3. Measured, not felt.
The short answer
THE_SMOOTHER 2 1.5.1 against Soothe 3 1.0.5, on 12 measured tests: ahead on 6, level on 5, behind on 1.
- Ahead: it removes more of the resonance at default settings, reacts sooner to a sudden resonance, leaves the sound between rings alone, leaves a held note alone, runs faster at 96 kHz, and is steadier when each plug-in is at its best setting.
- Level: resonance removed at each one's best setting, punch, pre-ring, latency, and CPU at 48 kHz on one of two mixes.
- Behind: steadiness as loaded on vocals and mixes, where Soothe 3 wobbles less.
THE_SMOOTHER 2 vs Soothe 3: the full scoreboard
| What you care about | THE_SMOOTHER 2 1.5.1 | Soothe 3 1.0.5 | Result |
|---|---|---|---|
| Removes the resonance, default settings (fix score, higher is better) | 21.1 | 15.8 | Ahead |
| Reacts to a sudden resonance (cut reached 25 to 50 ms in, same depth) | 10.5 dB | 3.9 dB | Ahead |
| Leaves the rest alone (level lost between rings that are cut by 9 dB) | 0.0 to 0.1 dB | 0.5 to 1.5 dB | Ahead |
| Leaves a single held note alone (6 kHz tone) | 0.9 dB cut | 7.3 dB cut | Ahead |
| CPU at 96 kHz (times faster than real time) | 71× | 49× | Ahead |
| Steady, no wobble, each at its best setting (lower is better) | 0.05 to 0.10 | 0.07 to 0.11 | Ahead on 5 of 6 jobs, level on 1 |
| Removes the resonance, each at its best setting (fix score) | 31.4 | 29.5 | Level, inside the measurement noise |
| Keeps the punch (drum hit level, same amount of processing) | 0.8 dB lost | 0.8 dB lost | Level |
| No pre-ring before a hit | −139 dB | −140 dB | Level, both inaudible |
| Latency at 48 kHz | 49 ms | 48 ms | Level |
| CPU at 48 kHz (times faster than real time) | 114× to 125× | 122× to 124× | Level on one mix, 7% behind on another |
| Steady, no wobble, as loaded (lower is better) | 0.13 to 0.22 | 0.12 to 0.24 | Behind on vocals and mixes, ahead on drums, level on low mids |
What the numbers mean:
- Fix score: I add two known resonances to a clean recording, run each plug-in on it, and score how close the result gets back to the clean recording. 0 means it did nothing useful, 100 means it removed the resonances perfectly and touched nothing else. It punishes leaving the resonance in and cutting anything that wasn't the resonance, so doing less or doing more never wins by itself. Measured over 24 clips across six jobs, once with the resonance on the whole time and once only on the louder half of the take, so a static EQ can't follow it.
- Default settings: both plug-ins confined to the job's frequency range, depth set to whatever scores best for each plug-in on each clip, every other control where it loads.
- Reaction: how many dB of cut each plug-in has reached at a given time after a resonance suddenly appears.
- Leaves the rest alone: both are set so the rings lose 9 dB; then I measure what the sound between the rings loses.
- Steadiness: how much the level wobbles under a held sound. A cut reacting to a hit is the job; a level moving while the music holds still is wobble. Lower is better.
- CPU: one instance, as times faster than real time. Higher is better.
How I made the test fair
Rule 1: measure Soothe 3 at its strongest. An early test of mine read Soothe 3 at one setting, and THE_SMOOTHER 2 looked great. Too great: one setting flatters whoever is compared against it. So for the scoreboard, Soothe 3 gets its best depth on every clip, and then several candidate settings in both its Soft and Hard modes, keeping its best on every job. THE_SMOOTHER 2 gets exactly the same treatment.
Rule 2: test on audio I never tuned on. Tune a plug-in on the same vocal you test it on and it will win on that vocal. That's a plug-in that memorised the exam. During development every change was checked on audio I never touched until it was time to judge. And the candidate settings for the scoreboard were searched on one job only, vocal harshness, so the other five jobs are new to both plug-ins. A change only counts if it holds on audio it has never heard.
Rule 3: no reward for doing less, or more. A plug-in that takes more of everything will always "remove more resonance", and it will also eat your vocal. The fix score compares the result with the clean recording, so over-cutting costs points just like under-cutting. And before I compare punch, steadiness or what's left around a ring, both are set to do the same amount of processing.
Both run as installed VST3 plug-ins in the same host, on the same audio. Each plug-in's latency is compensated, so the outputs line up with the input.
How I built THE_SMOOTHER 2 to beat Soothe 3
Step 1: a new engine that catches first and fixes second
The heart of THE_SMOOTHER 2 1.5.1 is a new engine with two stages working together. A fast stage catches a ring the moment it appears. A fine stage then takes it over with a precise notch, and the fast stage lets go. That's why it's on a sudden resonance so much sooner than Soothe 3:
| Time after the resonance starts | THE_SMOOTHER 2 | Soothe 3 |
|---|---|---|
| First 10 ms | 1.1 dB | 1.0 dB |
| 10 to 25 ms | 6.2 dB | 2.5 dB |
| 25 to 50 ms | 10.5 dB | 3.9 dB |
Both start gently, which keeps a drum hit intact. Then THE_SMOOTHER 2 is on it.
Step 2: a note is not a ring
The oldest complaint about resonance suppressors: they take the good stuff with the bad, because a held note looks a lot like a ring. The new engine tells them apart by how steadily they hold, and a tone with nothing around it, a synth line or a sine bass, gets the gentle treatment of a held note, even while it glides.
Early in development a slow sine sweep from 20 Hz to 20 kHz still lost 10.3 dB at its worst point. In 1.5.1 it loses 1.2 dB. Soothe 3 takes 7.5.
| A tone on its own | THE_SMOOTHER 2 | Soothe 3 |
|---|---|---|
| Held 100 Hz note | 1.0 dB cut | 1.5 dB cut |
| Held 1 kHz note | 0.9 dB cut | 2.3 dB cut |
| Held 6 kHz note | 0.9 dB cut | 7.3 dB cut |
| Slow sweep, 20 Hz to 20 kHz, worst point | 1.2 dB cut | 7.5 dB cut |
And between the rings, the part of your sound that was never a problem: set both so the rings lose 9 dB, and THE_SMOOTHER 2 takes 0.0 to 0.1 dB of what's around them. Soothe 3 takes 0.5 to 1.5 dB.
That's the "my track got thinner" feeling, measured. And gone.
Step 3: let the quiet passages breathe
Comparing the two outputs side by side on the same audio, one thing stood out: Soothe 3 left the quiet passages alone. Mine cut hardest exactly there, where you'd hear it most.
So I gave THE_SMOOTHER 2 its own way to handle it: it now follows how loud the band is right now, works on the loud moments that need it and lets quiet passages breathe.
I track progress with a band test: both plug-ins on one band, on audio they've never heard, 138 head-to-head checks. Over the development of 1.5.1:
| Stage | Wins | Ties | Losses |
|---|---|---|---|
| Start of development | 41 | 77 | 20 |
| Letting go faster after a ring | 47 | 76 | 15 |
| Following the band's loudness | 53 | 71 | 14 |
Step 4: calm lows
Low mids were the hardest job. Bass notes and low chords move constantly, and a cut that chases each one makes the low end wobble. Now, below about 1 kHz, the cut is averaged across frequency, so a broad low cut holds still while the bass moves through it. The low-mid fix score at default settings went from 13.9 to 19.3. Soothe 3 scores 21.2 there. That gap is inside the measurement noise: the low end is now as good as the one everyone calls the best.
Step 5: the last fix before release
One stage was letting go twice as fast as the RELEASE knob said. Fixing it so it releases at exactly the time you see lifted the default-settings fix score on every job, from 20.0 to 21.1. That's the 21.1 against Soothe 3's 15.8 in the scoreboard.
Over the 24 clips that lead is 5.3 points, with a 95% range of +1.9 to +9.5. It's a real lead, not luck: 16 clips go THE_SMOOTHER 2's way, none are tied, 8 go Soothe 3's.
Set each one up at its best and the gap closes: 31.4 against 29.5, a lead of 1.9 with a 95% range of −1.1 to +5.1 (13 clips won, 3 tied, 8 lost). Even with Soothe 3 tuned to its very best, it doesn't pull ahead. And at default settings, THE_SMOOTHER 2 wins by 5.3 points.
Where THE_SMOOTHER 2 and Soothe 3 are level
- Punch. At the same amount of processing, a drum hit loses 0.8 dB with both.
- Pre-ring. Nothing arrives ahead of a hit with either: −139 dB and −140 dB. You will never hear that.
- Latency. 49 ms against 48 ms at 48 kHz. Your DAW compensates both. THE_SMOOTHER 2 also has a LIVE mode at zero latency.
- CPU at 48 kHz. 114× to 125× real time against 122× to 124×. Level on one mix, 7% behind on another. At 96 kHz THE_SMOOTHER 2 pulls ahead, 71× against 49×.
The one row Soothe 3 takes
Steadiness, as loaded. When the music holds still, the level under it should hold still too. Lower is better:
| Job | As loaded: THE_SMOOTHER 2 | As loaded: Soothe 3 | Best setting: THE_SMOOTHER 2 | Best setting: Soothe 3 |
|---|---|---|---|---|
| Vocal harshness | 0.19 | 0.15 | 0.10 | 0.10 |
| Vocal air | 0.16 | 0.13 | 0.05 | 0.09 |
| Mixes | 0.22 | 0.17 | 0.06 | 0.09 |
| Drums | 0.15 | 0.24 | 0.06 | 0.11 |
| Low mids | 0.13 | 0.12 | 0.07 | 0.09 |
| Guitars | not comparable | not comparable | 0.05 | 0.07 |
With every control at its default, Soothe 3 holds a vocal and a full mix steadier. On drums THE_SMOOTHER 2 does, and low mids are level. Guitars as loaded are left out, because Soothe 3 didn't reach the same amount of processing there.
That's a trade I made on purpose: letting go of a ring faster means a cut no longer hangs into the next hit, and that costs a little steadiness at default settings. Set each plug-in up for the job and it turns around: THE_SMOOTHER 2 is steadier on five of six jobs and level on the sixth.
Why I measured against Soothe 3
Beating a weak plug-in proves nothing. So I went straight for the one most producers call the best, measured it at its strongest, and put every number on this page: 6 ahead, 5 level, 1 behind.
THE_SMOOTHER 2 beats every resonance suppressor I've tested. Soothe 3 isn't the only one. Here's what happened to an AI-made resonance suppressor on the same kind of test: it barely found the rings, while THE_SMOOTHER 2 took all four down and left the sound between them untouched.
Think another resonance suppressor deserves the title? Tell me which one. It goes through the same test, and the result goes on this page.
Run the test yourself
Every signal in this test is one you can rebuild in ten minutes: pink noise with a few rings on top, a held sine, a slow sweep, clicks on silence. Put Soothe 3, or any resonance suppressor you own, next to THE_SMOOTHER 2, set both to the same amount of processing and look at what's left.
Or do it on your own vocal. Hit DELTA and you hear exactly what THE_SMOOTHER 2 takes out, and nothing else.
Takes the ring. Leaves the note. At a fraction of the price.
How I measured this
- Material
- Fix score: 24 job-clip pairs from 19 clean recordings across six jobs: vocal harshness 2-6 kHz (6 clips), vocal air 4.5-10 kHz (5), drum loops 5-14 kHz (3), guitars 2-5 kHz (2), mixes 2-6 kHz (3), low mids 200-500 Hz (5). The loudest 12 s of each clip get two resonances added (peaking filters, +9 dB, Q 10, at 24% and 67.5% of the way through the job's band on a log scale), in two scenarios: on the whole time, and faded in only on the louder half of the take so a static EQ cannot follow it
- Other rows: test signals anyone can rebuild: pink noise with four rings 18 dB over it, held sine tones at -12 dBFS, a 20 Hz to 20 kHz sine sweep over 20 s, and clicks on silence
- Versions
- THE_SMOOTHER 2 1.5.1 by Phil Speiser
- Soothe 3 (soothe3) 1.0.5 by oeksound, $259 list price at the time of the test
- Measured on
- 2026-11
- Settings
- Fix score, default settings: every control at its default except two. Both plug-ins are confined to the job's band (THE_SMOOTHER 2: one band set to the range; Soothe 3: steep low-cut and high-cut bands at the range edges), and DEPTH is set per clip and scenario to the value that scores best for that plug-in (golden-section search, 8 renders). Soothe 3 defaults: Soft, Detail 5, Attack 5, Release 3, Quality normal
- Fix score, best setting: per job, each plug-in's highest-scoring candidate setting, depth again set per clip. Candidates came from a random search on the vocal harshness job only, so the other five jobs are out of sample. THE_SMOOTHER 2 had six candidates (including its defaults and two hand settings), Soothe 3 had five (its defaults plus Hard and Soft settings at high or ultra quality)
- Fix score formula: output and clean recording compared in 1/12-octave by 20 ms cells inside the job's band (cells within 45 dB of the clean take's loudest). Error = RMS level difference in dB. Score = 100 x (1 - error / error with the plug-in bypassed): 0 = no improvement, 100 = resonances removed perfectly and nothing else touched, negative = worse than bypass. A pair's score is the mean of its two scenarios; the headline is the mean over the 24 pairs. 95% range: bootstrap over the 24 paired differences, 20,000 resamples. A clip is won or lost when the difference exceeds 0.5 points
- Reaction, punch, steadiness and what is left alone are compared with both set to the same amount of processing
- Held constant
- Both run as installed VST3 plug-ins in the same host, on the same audio, at the same sample rate and buffer size. Each plug-in's reported latency is compensated, so the outputs line up with the input
- Notes
- Measured on Windows at 48 kHz. CPU is one instance on the same machine, as times faster than real time. Guitars as loaded are left out of the steadiness row because Soothe 3 did not reach the same amount of processing there. The development figures (the 138-check band test, the lone sweep, the low-mid and as-loaded score steps) come from the THE_SMOOTHER 2 1.5.1 development log
FAQ
Is THE_SMOOTHER 2 better than Soothe 3?
Measured side by side on 12 tests, THE_SMOOTHER 2 1.5.1 is ahead of Soothe 3 1.0.5 by oeksound on 6, level on 5 and behind on 1. At default settings, with each plug-in's depth set to its best value, it removes more of the resonance (fix score 21.1 against 15.8 over 24 clips, 95% range of the lead +1.9 to +9.5), reacts sooner to a sudden resonance (10.5 dB of cut against 3.9 dB, 25 to 50 ms in), leaves the sound between rings alone (0.0 to 0.1 dB lost against 0.5 to 1.5 dB) and cuts a held 6 kHz note by 0.9 dB against 7.3 dB. Soothe 3 is steadier as loaded on vocals and mixes.
Is THE_SMOOTHER 2 a good alternative to Soothe 3?
Yes, on the measurements. THE_SMOOTHER 2 1.5.1 matches or beats Soothe 3 1.0.5 on 11 of 12 measured tests, at a fraction of Soothe 3's $259 list price. With each plug-in at its best setting both remove the same amount of resonance (fix score 31.4 against 29.5, inside the measurement noise), and at default settings THE_SMOOTHER 2 removes more (21.1 against 15.8).
What is the fix score in the THE_SMOOTHER 2 vs Soothe 3 test?
Two known resonances are added to a clean recording, each plug-in processes it, and the score says how close the result gets back to the clean recording: 0 means no improvement over bypass, 100 means the resonances are removed perfectly and nothing else is touched, and a negative score means worse than bypass. Because it compares against the clean recording, it punishes both leaving the resonance in and cutting anything that was not the resonance, so it does not reward doing less or doing more. It is measured over 24 clips across six jobs, with the resonance on the whole time and only on the louder half of the take.
Where is Soothe 3 better than THE_SMOOTHER 2?
On steadiness as loaded. With every control at its default, the level under a held vocal or a full mix wobbles a little less with Soothe 3: 0.15 against 0.19 on vocal harshness, 0.13 against 0.16 on vocal air and 0.17 against 0.22 on mixes, lower being better. On drums THE_SMOOTHER 2 is steadier (0.15 against 0.24) and low mids are level. With each plug-in at its best setting THE_SMOOTHER 2 is steadier on five of six jobs and level on one. At 48 kHz THE_SMOOTHER 2 is also 7% slower than Soothe 3 on one of two mixes.
Which reacts faster to a sudden resonance, THE_SMOOTHER 2 or Soothe 3?
THE_SMOOTHER 2. In the first 10 ms both are level, at 1.1 and 1.0 dB of cut. From 10 to 25 ms THE_SMOOTHER 2 has reached 6.2 dB against Soothe 3's 2.5 dB, and from 25 to 50 ms 10.5 dB against 3.9 dB, both set to the same depth.
Does Soothe 3 or THE_SMOOTHER 2 cut held notes?
Soothe 3 cuts a held note more. A held 6 kHz tone on its own loses 7.3 dB with Soothe 3 1.0.5 and 0.9 dB with THE_SMOOTHER 2 1.5.1. At 1 kHz it is 2.3 against 0.9 dB, and at 100 Hz 1.5 against 1.0 dB. A slow sine sweep from 20 Hz to 20 kHz loses 7.5 dB at its worst point with Soothe 3 and 1.2 dB with THE_SMOOTHER 2.
How do THE_SMOOTHER 2 and Soothe 3 compare on CPU and latency?
At 96 kHz one instance of THE_SMOOTHER 2 runs 71 times faster than real time against Soothe 3's 49 times. At 48 kHz they are level on one mix (114 to 125 times against 122 to 124 times) and THE_SMOOTHER 2 is 7% behind on another. Latency at 48 kHz is 49 ms for THE_SMOOTHER 2 and 48 ms for Soothe 3, and a DAW compensates both. THE_SMOOTHER 2 also has a LIVE mode with zero latency.
Is Soothe 3 the best resonance suppressor?
Soothe 3 by oeksound is the resonance suppressor most producers name as the best, and it is a very good plug-in. In a side-by-side measurement on 12 tests, THE_SMOOTHER 2 1.5.1 is ahead of it on 6, level on 5 and behind on 1, and THE_SMOOTHER 2 beats every resonance suppressor Phil Speiser has tested.
How was THE_SMOOTHER 2 compared with Soothe 3?
Both ran as installed VST3 plug-ins in the same host on the same audio at 48 kHz, with each plug-in's latency compensated. For the fix score both were confined to each job's frequency band and their depth was set per clip to its best value, first with every other control at its default, then at each plug-in's best candidate setting per job (six candidates for THE_SMOOTHER 2, five for Soothe 3, in both its Soft and Hard modes). Wherever a side effect is compared, both were first set to the same amount of processing.
