VOCANT

1. Introduction

Most effects plug-ins give you a hundred knobs and leave you to work out which twelve matter for the sound in your head. Vocant turns that around. You describe the sound, and it builds the chain.

Type warm vocal with plate reverb and light compression and press Enter. A few seconds later a compressor and a plate reverb are active, in signal-flow order, with their controls set to values that make sense for a vocal. Type add some chorus and a chorus joins them without disturbing what is already there. Every control is still a knob you can turn, so the result is a starting point you own, not a black box.

How it works, in one paragraph

Vocant is two programs. The plug-in is a VST3 with 32 effect modules and 303 parameters, all of them ordinary automatable parameters as far as your DAW is concerned. The companion is a small service the plug-in launches for you. It runs a language model on your machine, reads your prompt plus the state of your chain, and answers with a list of changes. Nothing about your audio ever leaves the plug-in; the model only ever sees text.

What it is for

• Getting a usable chain on a track in seconds, then refining by ear.
• Guitar and bass rigs built from your own Neural Amp Modeler captures and cabinet impulse responses, chosen by name.
• Vocals, drums, synths, anything you can describe in the language you would use to another engineer.
• Learning. Ask for a sound, then look at what it did.

2. Installing

You will need

Running the installer

1. Run Vocant-Setup-1.0.0.exe. It installs per user, so no administrator rights are needed.
2. Accept the licence agreement.
3. Choose where the companion lives. The default is inside your local application data folder; pick another drive if your C: drive is tight, because the companion plus its model needs about 3.5 GB.
4. Finish. The plug-in is now in your user VST3 folder and your DAW will find it on its next scan.

Checking what you downloaded

Every release is published with a SHA-256 hash. Before you run the installer, open PowerShell in the download folder and compare:

Get-FileHash .\Vocant-Setup-1.0.0.exe -Algorithm SHA256

The hash it prints must match the published one exactly. If it does not, delete the file and download again.

What the installer touches

• The plug-in bundle, in your user VST3 folder (%LOCALAPPDATA%\Programs\Common\VST3).
• The companion, in the folder you chose.
• One registry value under HKCU\Software\Vocant so the plug-in can find the companion.
• A Start-menu shortcut for launching the companion by hand, if you ever want to watch its console.

On the first launch of the plug-in, the companion downloads the language model, about 3.3 GB, from Hugging Face. That is the only time Vocant uses the network unless you opt into cloud inference (section 8).

Where it goes in the chain

Vocant is a complete rack, so it usually goes where you would put the whole rack: as an insert on the track, after any tracking-stage processing you always want and before the bus. Its own modules are already in a sensible order, dynamics first and time-based effects last, and you can drag them into any other order.

3. Quick Start

1. Insert Vocant on an audio track that has something playing on it.
2. Watch the top right of the plug-in. The dot next to OSC fills in and a backend badge such as VULKAN or CUDA lights when the companion is up. On a brand-new install this takes as long as the model download; after that, a few seconds.
3. Click the prompt box in the header, type big hall reverb, and press Enter.
4. The REV module lights in the rack with a small flag, its screen opens, and the controls move to the model's choices. You will hear the reverb on your audio.
5. Turn any knob. Everything the model set is yours to change.

Build a chain

Now try a description rather than an effect name:

warm vocal with plate reverb and light compression

The compressor and the reverb both activate, and the reverb switches to the Plate algorithm. Then extend it:

add some chorus

Chorus joins the chain. The compressor and reverb stay exactly where they were. Section 4 explains why the wording matters.

4. Writing Prompts

The model understands audio-production language. Describe the sound the way you would describe it to another engineer, and let it pick the modules.

Describe the sound, not the signal path

Name the source

The same adjective means different settings on different material. warm vocal gets a presence boost, moderate compression and a plate; warm bass gets low-end weight, compression and a little saturation. Say what you are processing.

Adjectives the model maps well

Exact values are honoured

Quantities in the prompt are applied literally rather than approximated. 200 ms delay, -6 dB, 1/8 note and high-pass at 80 Hz land as typed. If a prompt names nothing the chain can act on, Vocant shows a notice rather than an error and leaves the chain alone.

Add to the chain, or replace it

Every prompt is sent together with the list of modules already active, so the model knows what is there. Two phrasings get two behaviours:

Iterate

1. Start broad: rock guitar.
2. Listen, then steer: more saturation, less reverb.
3. Extend: add a slapback delay.

Repeat prompts return instantly because results are cached for the session.

5. The Rack and the Modules

The strip of icons under the header is the rack. It runs in signal order from INPUT on the left to OUTPUT on the right. A lit icon is an active module; a small flag on an icon means the last prompt changed it. Click any icon to open that module's screen below the rack.

• Reorder by dragging icons along the rack. The order is the signal path.
• Bypass a module with the BYPASS toggle in its screen header, to compare with and without it.
• Turn a knob by dragging. Double-click its readout to type an exact value.
• The DK/LT button in the header switches between the dark and light themes. The label names the theme you would switch to.
• The gear at the far right opens Settings: capture folders, inference provider, and activation.

Every module has the same shape

A module screen is a red-framed main panel and, for modules with output controls, a separate neutral OUTPUT panel to its right holding mix, width and level. Controls sit on identical rows across every module, so once you have read one screen you can read them all.

The modules

Input and Output

The INPUT screen has an input gain, meters and a live spectrum. The OUTPUT screen carries the master section: a true-peak safety limiter at -1 dBTP with 1.5 ms of lookahead (reported to your DAW as latency), and Auto-Normalize, a static output trim aimed at -3 dB that re-aims only when the chain changes, never while you play. It is gain staging, not an automatic gain control.

Tempo

Modulation and delay modules have a SYNC toggle and a DIV dropdown. With sync on, the rate or time follows your DAW's tempo at the chosen note division.

The header, left to right

6. Amp Captures and Cabinets

Vocant plays Neural Amp Modeler captures, the .nam files, and cabinet impulse responses, in three slots: NAM PEDAL for stompbox and preamp captures, NAM AMP for amplifiers, and IR CAB for speaker cabinets. Vocant ships no captures. It plays the ones you own.

Telling Vocant where your files are

1. Open Settings with the gear and choose the NAM PATHS tab.
2. Add a folder under each category you use. The categories matter:

3. Click SAVE & RESCAN. Folders are scanned recursively, so one top-level folder is enough.

Loading by hand

Open the slot's screen and pick a file from the MODEL dropdown. RESCAN picks up files added since the last scan. The status line under the dropdown reads MODEL ENGAGED once the capture is running on the audio thread.

Loading by name

Once folders are set, the model can see your filenames and choose among them. Fender clean with spring verb loads the nearest Fender capture and a spring reverb; Mesa Rectifier modern metal loads a Mesa capture and a matching 4x12 response. If a word in your prompt matches a filename you own, that file is loaded even when the model would have picked another.

Three phrases are treated as explicit requests for an amp tone and put your whole amp inventory in front of the model: electric guitar, electric bass, and bass on its own. Plain guitar does not, so an acoustic guitar prompt never pulls in an amp, and EQ language such as boost the bass goes to the EQ as it should.

The NAM Amp screen

• INPUT drives the capture. Captures trained at a low input level can need this pulled well down before they clean up.
• GATE and GATE REL are a noise gate that triggers ahead of the drive, so its threshold does not move when you change the input.
• BASS, MID and TREBLE are a modelled Bassman tone stack. TONE STACK engages it and TS PRE moves it in front of the amp.
• OUT NORM levels the capture's output so different files sit at a similar loudness. IN NORM is opt-in input calibration; leave it off unless a capture asks for it.

7. Presets and Sharing

Vocant has no preset browser of its own. The prompt box is the browser. Saving and recalling a state is your DAW's job, and every DAW does it the same way for any plug-in: the whole chain, every parameter and the module order, as one preset.

Sharing a sound

The simplest way is to share the prompt. Another Vocant user who types the same words gets the same character, if not the identical numbers. For an exact copy, export your DAW's preset file and send that.

Starting over

In REAPER, pick Reset to factory default from the preset list. In any DAW, removing and re-adding the plug-in does the same.

8. Cloud Inference (Optional)

Out of the box, the model runs on your machine and nothing leaves it. If you would rather have faster or sharper answers, you can point the companion at a cloud provider. This is opt-in, per provider, and reversible.

1. Open Settings and choose the LLM INFERENCE tab.
2. Pick a provider. The key field appears.
3. Paste the key and click SAVE & CONNECT. The key goes into Windows Credential Manager, never to a plain file, and the companion reconnects in the background. No restart.

The panel shows a running cost for the session. Requests cost a fraction of a cent each. To go back, choose LOCAL.

9. Licensing and Privacy

Trial

Vocant runs as a full-featured 14-day trial from the first time it is loaded. Every module, every parameter and every prompt is available. A banner in the plug-in shows the days remaining.

When the trial ends, the plug-in mutes its audio and the meters go dark until it is activated. Your sessions are safe: the state is kept, and activating restores everything exactly as it was.

Activating

1. Buy a licence at reclusestudios.net. Your key arrives by email.
2. Open Settings, choose the ACTIVATION tab, or use the form in the trial banner.
3. Enter the email address and the key, and click ACTIVATE. The plug-in contacts the licence server once, stores the result locally, and does not need the network again.

Privacy

• Audio never leaves the plug-in. The companion receives text only.
• Prompts stay on your machine unless you enable cloud inference (section 8).
• The plug-in and companion talk over UDP on 127.0.0.1 only.
• Network use is limited to the one-time model download, activation, and cloud inference if you opt in. Nothing phones home.

10. Troubleshooting

11. Specifications

12. Glossary