Meet New VST Plugins: Speech Cleaner & Voice Dereverb

LALAL.AI’s plugin family has grown. Alongside Stem Separator, our VST plugin that isolates vocals, drums, piano, bass, and both electric and acoustic guitars right inside a DAW, there are two new arrivals built around the human voice: Speech Cleaner and Voice Dereverb. One lifts speech out of background noise, and the other draws the room out of a voice recording.

Both plugins are now available to LALAL.AI Pro users, so let’s take a look at each of them. 

Speech Cleaner Pulls Dialogue out of the Noise

Air-conditioner hum under an interview, traffic bleeding through the window in a vlog, and any other case that gets your voice lost in background noise is precisely the kind of mess Speech Cleaner was made for. The name says as much, implying upfront what material the plugin expects, as it works only with speech content (stereo tracks).

Noise Level control comes down to a single slider, running from 0% to 100%, with Clean at the left end and Original at the right. Left at its default of 0%, the slider delivers the cleaned speech alone. If you raise it, a measure of the original noise finds its way back, which often makes dialogue sound more like it was recorded somewhere real. You’ll find it useful if you’d like to keep some environmental ambience and maintain the original feel of each scene for authentic localization or dubbing.

Any value above 100% amplifies the noise by 0.06 dB per percent and can only be typed.

You can also control Latency by moving the dedicated slider. 

As an output, you get three tracks:

• Main with the mix set by Noise Level 
• Voice with the cleaned speech alone 
• Noise with everything the model removed (input without the cleaned speech)

Use the optional Voice and Noise outputs to balance them on separate mixer channels or to process the noise bed separately. Note that the Voice and Noise outputs are disabled by default and need to be enabled in your host’s multi-output routing. The Noise Level slider doesn’t affect them.

Voice Dereverb Takes the Room out of the Recording

Few things give away a home recording faster than the room it was made in. Voice Dereverb goes after exactly that, stripping echo and reverb tails from a voice and leaving it drier and more present than the space allowed.

One critical condition comes attached, though. The plugin is built for clean voice material (stereo only), such as studio recordings, solo vocals, clear speech, or the output of Speech Cleaner. 

Note that if you hand it a voice buried in a full mix, the expected result can’t be guaranteed, since the model was never trained on such input. 

Reverb Level spans 0% to 200%, with Original sitting at the center. At 0% (default) the voice comes out completely dry. Anywhere between 0% and 100%, the room is dialed back to taste, and at exactly 100% the signal passes through as recorded. 

Anywhere above 100% the reverb grows louder than in the original recording and goes further still when a value up to 999% is typed in. Exaggerating a room is a perfectly legitimate creative move, and here it’s treated as one.

As with Speech Cleaner, two optional buses accompany the main output:

• Main with the mix set by Reverb Level 
• DryVoice with the de-reverberated voice alone 
• ReverbTail with everything the model removed (input without dry voice)

And again just as with Speech Cleaner, you can use the optional outputs to balance voice and room on separate mixer channels or to treat the reverb tail with its own processing. 

From the Recording Studio Right to Your Edit Suite

Noisy, roomy dialogue is hardly a problem unique to music, and nowhere is it more familiar than in video post-production. Location sound rarely arrives clear, and the fix is usually needed inside the editor rather than in a separate audio application. That is why both plugins have been tested and confirmed to work in DaVinci Resolve.

On Windows and Linux, the plugins ship as VST3. On macOS, a single universal binary for Apple Silicon and Intel offers both VST3 and AU and requires macOS 13.3 or later. 

Since each plugin does one job, the two combine naturally. For a recording that is both noisy and echoey, put Speech Cleaner first in the chain and Voice Dereverb after it. The first lifts the voice out of the noise, and the second dries out what is left of the room, handing the mix a far cleaner starting point than the raw take.

Three Plugins, One Shared License

Like Stem Separator before them, both newcomers process audio locally, on the user’s own hardware, with nothing sent to the cloud. Each plugin picks the fastest backend available when it loads, using the GPU where it can and falling back to the CPU where it cannot. An internet connection is needed only for the license activation and occasional subscription checks.

No separate purchase is involved. A LALAL.AI Pro subscription unlocks the whole plugin lineup on up to three devices, and activating any plugin on a machine activates the rest along with it. 

To get started, you need to download Speech Cleaner and Voice Dereverb from the LALAL.AI plugin page, install them on your computer, activate them, and restart your DAW or DaVinci Resolve. In most cases, the plugins will appear among the available tools right away; if not, run a plugin rescan.

How to Install the Plugins

Here’s a quick overview for each platform. Every plugin comes with its own installer, so you only need to install the one you plan to use.

How to Activate the Plugins

Activation works the same way on every platform. You’ll need a LALAL.AI Pro subscription and an internet connection.

1. Open the plugin in your DAW or video editing software and click Activate.
2. Read the EULA and click Accept. It appears only once per device.
3. Enter the email of your LALAL.AI account and click Log In.
4. Check your inbox for a one-time code and enter the 6-digit code in the plugin. If the email hasn’t arrived, check your spam folder or click Resend.
5. Once the code is verified, the plugin unlocks.

FAQ 

Do LALAL.AI plugins work with mono tracks? 

Mono tracks aren’t supported. Insert the plugin on a stereo track, or let the host upmix a mono track to stereo.

Which sample rates are supported? 

Any host sample rate is accepted. The models run at 44.1 kHz internally and the plugin resamples on the way in and out.

Do I need a LALAL.AI Pro subscription to use the plugins? 

Yes, all three plugins are only available to those LALAL.AI users who have an active Pro (Monthly or Yearly) subscription. There’s no free trial at this point. One subscription includes all LALAL.AI plugins and lets you use them on up to three workstations. A workstation counts once no matter how many plugins are installed on it. Activations are personal and can’t be shared with other users.

Do I need an internet connection to work with the plugins? 

Internet access is only required for occasional subscription checks. Audio processing itself runs entirely offline.

What if I hear glitches or dropouts when working with the plugins? 

Each plugin needs a buffer of audio before its model can produce output, and it reports that delay to the host so plugin delay compensation keeps the track in sync. When you move the latency slider and click Apply, you should expect a brief glitch while the pipeline refills; a few DAWs also need a restart to pick up the new delay compensation.

If you hear glitches or dropouts:

1. Increase latency (at least to the value named in the overload banner) and hit Apply.

2. Reduce system load, for example by increasing your DAW’s buffer size or by freezing tracks with other heavy plugins.

3. Resume playback and recheck stability.

Does changing the latency affect audio quality? 

In Stem Separator and Voice Dereverb, calibrating the latency never affects the quality of audio output. 

In Speech Cleaner, a lower latency can slightly affect quality, but only during playback. When you export or render a track with the plugin, for example with bounce in place, Speech Cleaner always uses the maximum latency and delivers its best quality.

Why can’t I control latency in DaVinci Resolve? 

DaVinci Resolve (just like Audacity) pre-renders faster than realtime, so the plugins always run in the render path at maximum latency. This is why the latency control is disabled there. In Resolve, control changes reach the host a moment after you click, once Resolve has finished re-rendering the previous change.

What if the plugin doesn’t appear in my DAW?

Most DAWs scan the default VST3 and AU folders at startup. If the plugin still isn’t listed:

1. Open your DAW’s plugin preferences.
2. Make sure the plugin folder is included in the scan list.
3. Run a full plugin rescan.
4. Restart the DAW once more, as some hosts require it.

Are minutes deducted from my account when working with the plugins?

No. The plugins process audio locally, so no minutes are deducted from your account.

Follow LALAL.AI on Instagram, Facebook, Twitter, TikTok, Reddit, LinkedIn, and YouTube to keep up with all our updates and special offers.