V-NAM

What it is

V-NAM is a Fwuzeem Audio Neural Amp Modeler host for one .nam capture. The selected capture is loaded into two independent NAM instances, one for each channel, so the processing is dual-mono rather than a single shared stereo state. Input and output staging use the V-Shift exponent law, with optional sample-rate render modes, L/R or M/S routing, latency-aligned dry mixing, and Live or Render engine profiles.

Controls & parameters

Editor detail levels

The editor has Simple, Standard, and Full detail levels. The selected level is stored as the guiDetail state property; its default is 1 (Standard). Simple keeps the core shift, Wet/Dry, Bypass, and model controls visible; Standard adds Link Shifts and the information strip; Full adds Retain Sample Rate, Oversample, Stereo Mode, Engine Mode, and Precision. Input Gain is shown only when Link Shifts is on and the detail level is Standard or Full, although its parameter remains active and automatable at every level. The editor’s rotary controls use the house range pi*0.82 to pi*2.18 with a stop at each end.

The shift law

V-NAM’s shift controls are not ordinary dB knobs. Within the normal finite float range, a +1 shift multiplies each nonzero sample by 2 exactly. A -1 shift divides it by 2 exactly. One shift is 6.0206 dB; the range is 20 doublings or halvings in either direction. A +1 input shift followed by a -1 output shift therefore returns the staging ratio to unity. The same law is used by Input Gain and Output Shift. Zero and non-finite samples are skipped; extreme values are clamped by the float implementation, and denormals can flush to zero. The DSP applies shifts without smoothing.

Workflow / how to use it

1. Insert V-NAM on a mono or stereo track. The input and output layouts must match.
2. Choose LOAD .NAM and select a capture. V-NAM loads the file as a separate NAM state for each channel. The previous and next buttons cycle through sorted sibling .nam files in the same directory; CLEAR removes the model.
3. Start with Input Shift, Input Gain, and Output Shift at 0 for unity staging around the capture. To drive the model by one exact doubling while keeping the output staging ratio unchanged, use +1 on Input Shift and -1 on Output Shift.
4. Use L/R for conventional dual-mono processing. Use M/S on a stereo source when the model should receive Mid and Side lanes; the plugin decodes them back to L/R afterward.
5. Choose the render behavior. Retain Sample Rate feeds the model at its captured rate. With Retain Sample Rate off, Off uses the host rate, while 2x and 4x use two or four times the model’s native rate. These are creative rate changes, not transparent anti-aliasing oversampling.
6. Use Live while tracking and Render when the generic Render engine is preferred for a bounce. Check the host’s reported latency after changing the model or rate mode.
7. Reduce Wet/Dry to blend the processed model with the latency-aligned dry signal. Use Bypass for a timing-matched dry comparison.
8. Save the project or plugin state to store the parameters and the model path.

Use cases

• Guitar or bass capture at unity staging: Load one capture and leave all three staging controls at 0. This gives a clear starting point for hearing the capture without adding an intentional input or output doubling.
• Doubling the model voicing: Set Oversample to 2x with Retain Sample Rate off. The model is fed at twice its native rate, which the editor describes as a one-octave-up voicing change. Use it as a creative model-rate choice, not as a transparent oversampling claim.
• M/S mic pair processing: Use M/S on a true stereo pair to send Mid and Side through the two independent channel states, then decode back to ordinary stereo.
• Link-shift gain automation: Engage Link Shifts for an inverse input/output pair while adjusting the editor’s shift control, so a harder model drive can be level-compensated by the opposite output shift. The editor callback can mirror a host-driven shift while the editor is open; for editor-independent automation or offline rendering, automate both shift parameters explicitly.
• Wet/dry model blending: Set Wet/Dry below 100 percent to retain a latency-aligned clean component while auditioning a capture’s saturation or nonlinear character.

Specs & requirements

• Brand: Fwuzeem Audio; manufacturer code Fwzm.
• Formats: AU and VST3.
• Audio I/O: Mono or stereo input/output, with matching layouts. Stereo processing uses two independent model instances. No MIDI processing is enabled.
• Sample rates: No fixed sample-rate list is declared by the plugin. A valid model sample rate supplies the native rate; an invalid or missing model rate falls back to the host rate.
• Render rates: Retain Sample Rate uses the model native rate. Otherwise, Off uses the host rate, and 2x/4x use native-rate multiples.
• Latency: With no model loaded, reported latency is 0. With a model loaded, latency comes from the rate-conversion container and varies with the host/native/render-rate relationship. The dry ring and Bypass path use the reported latency, but its delay request is clamped to 65,535 samples by the 65,536-sample ring, so alignment is not unconditional above that ceiling.
• Precision: The processor accepts FP32 and FP64 host callbacks. The Precision parameter is not a DSP switch.
• State: Parameters and the model path are stored as APVTS XML state. The plugin exposes one host program and does not implement preset programs.
• Platform: Workspace audio-toolbox build guidance targets macOS 10.15 or later.

Notes / caveats

• Precision is display-only. Host callback precision determines which process callback is used; selecting FP64 does not force a host to call the double-precision path.
• Link Shifts is editor-only. When the editor is open and Link Shifts is on, its slider callbacks mirror a host-driven change to the sibling shift parameter. With no editor open, the processor does not read linkShifts, so editor-independent automation should set both shift parameters explicitly.
• The parameter Wet/Dry supports 0.1 percent steps, while the current rotary editor displays a rounded whole percent.
• Oversample is a creative model-rate control. Its 2x and 4x rates are based on the model’s native rate, not necessarily the host rate. Retain Sample Rate overrides the Oversample choice.
• A missing model does not mute the plugin: the NAM stage is skipped, while input shift, Input Gain, M/S routing when applicable, Output Shift, and Wet/Dry still run. A failed load shows a warning and restores the previous model.
• On state restore, a missing saved model file leaves the last-known path in the state but leaves the engine unloaded. There is no file watcher in the plugin source; load the capture again after it is available.
• The host latency reported for a loaded model is the resampler’s value. Do not assume that the Off label means zero latency when a model is loaded.