Repository navigation
Releases
PNAM Player 0.5.5
Mono and stereo
- The button next to Output Gain shows a single ring in mono and two overlapping rings in stereo.
- Stereo processes the left and right inputs independently through the same selected NAM or PNAM model and IR. Parametric controls, model size, gain, and IR bypass remain shared.
- Mono uses the left input and sends the same processed signal to both outputs. Existing sessions default to mono. Stereo requires a host layout with two input channels.
Full Changelog: v0.5.4...v0.5.5
PNAM Player 0.5.4
PNAM Player 0.5.4
This release changes the product name from NAM Parametric Plugin to PNAM Player and identifies the vendor as The Tone Scientist (manufacturer code TTSc), matching NAM Oversampler's vendor grouping. The unique plug-in code is now PnPl, distinct from Phillip Self's upstream NpPg.
- The main GUI title PNAM Player is entirely red, the version is shown beneath it, and the red update indicator appears beside the version.
- Input Calibration uses an editable numeric field without a slider, with 0.001 dBu resolution and exactly three displayed decimal places. The editable value stays centered when focused; its background and the Output Mode selector use the dark model-parameter-area color. The parameter ID is unchanged.
- Fixed a bug on Output mode: input/output calibration values and gear type are now read from outer metadata, while loudness is selected from the active Lite/Full model with fallback to outer loudness.
- The neural audio engine, effects and model compatibility remain unchanged.
DAW session migration: Changing product name and manufacturer/plug-in IDs may make PNAM Player appear as a different plug-in from the previous release. Back up projects that use the older plugin before migrating.
Packages: Windows x64 VST3 and Standalone; macOS universal VST3, AU and Standalone; Linux x86_64 VST3 and Standalone. The seven packages are built without repeating previously completed model, calibration, audio, IR or export functional tests. The release stays draft and is not published.
NAM Parametric Plugin 0.5.3
NAM Parametric Plugin 0.5.3
macOS Standalone: restore CoreAudio input access
- Fix missing macOS microphone/audio-input privacy usage declaration. The standalone application's
Info.plistnow containsNSMicrophoneUsageDescriptionwith a clear explanation that access to the audio interface is needed to process a guitar or other instrument. This enables macOS to request and manage the appropriate permission when the standalone uses CoreAudio inputs. - Add a macOS build-time regression gate that checks the bundled app's
Info.plistfor the microphone permission text, the expected version and an unchanged JUCE default-mute wrapper.
Packages
Seven verified native packages, built from the same release commit:
- Windows x64: VST3 and Standalone
- macOS universal (Apple Silicon + Intel): VST3, AU and Standalone
- Linux x86_64: VST3 and Standalone
Windows/macOS packages are not signed or notarized. See the README for installation information.
Updated builds — 12 November 2026
The seven downloadable packages have been rebuilt and verified after moving the new-version notification LED beside the large Parametric NAM title. Update checking, blinking and the release-link click action are unchanged.
The updated binaries come from commit 489ae88a2f81. The original Git tag and auto-generated source archives remain unchanged.
SHA-256 of refreshed packages:
NAM-Parametric-Plugin-0.5.3-Linux-x86_64-Standalone.tar.gz:39b0e8c0b3bd9cc6cb03ab555d155c82bbe4110667aa230277f3a7bc967cf829NAM-Parametric-Plugin-0.5.3-Linux-x86_64-VST3.tar.gz:4c95a4f20bac77bc41e4c4e82917a16f96f849c2143086e706c5ec24d8cee9d4NAM-Parametric-Plugin-0.5.3-Windows-Standalone.zip:1628221f16970c23dc6ea70193a29c00e752f00a76491707ddfbe2538b81b0ebNAM-Parametric-Plugin-0.5.3-Windows-VST3.zip:5bde74157a3c940d345f6f0a731a0c5bae8365a91dfdae42ef4457385519738fNAM-Parametric-Plugin-0.5.3-macOS-AU.zip:a7292dd2885fc04fb41c40a4a1aec4d6a9166aea0ba6e73187a9d8a6e4d7a6baNAM-Parametric-Plugin-0.5.3-macOS-Standalone.zip:9f291a7f642ccbd80645b808277dbb7ca7702d57867dd2c8284ce1b566196babNAM-Parametric-Plugin-0.5.3-macOS-VST3.zip:0b4ed64d28f0ac9231f713eec2850f014c4520ff94b8dc5164b582fc71794fad
NAM Parametric Plugin 0.5.2
NAM Parametric Plugin 0.5.2
Version 0.5.2 combines faster parametric-control interpolation with a discreet, opt-in update notification in the plugin UI. The 100/100 ms Linear control behavior is unchanged.
Faster real-time weight smoothing
- During movement of an A2/HyperWaveNet control, convolution kernels now calculate the delta-weight contribution using preallocated Eigen matrix multiplications (GEMM) instead of per-coefficient scalar products. The intermediate coefficients remain mathematically equivalent to the former per-sample Linear interpolation.
- Reuses one neural audio network, with at most one hypernetwork update every 100 ms and a 100 ms Linear coefficient ramp at the audio sample rate.
- An unmodified convolution skips the delta work entirely. The fast stationary convolution path remains intact.
- All scratch arrays are allocated before audio processing. This optimization does not use the PR's optional multithreaded/off-audio-thread hypernetwork API, additional audio DSP instances or crossfades.
- NeuralAudio's reference A2 implementation was tested in an isolated A/B/C run using the same inputs and sample-accurate automation: no meaningful audio regressions, with significantly lower smoothing CPU cost. The plugin uses its own pinned parametric core, so native plugin CI and real capture listening remain important.
Quiet notification of newer versions
- The plugin checks the official
DLC86/NamParametricPluginGitHub releases asynchronously on its UI background task. - When a published version newer than the installed version is found, a small blinking red LED appears beside the Settings button.
- Clicking the LED opens that version's official GitHub release page.
- No popup or startup message, no automatic update or download; the LED remains hidden if there is nothing newer or the network is unavailable. No network operation occurs in the audio callback.
- Published version metadata is checked using numeric semantic-version comparison. Drafts are ignored.
Still supported
- Legacy
enum_namesswitches and theoptions/name/value/index_swformat without modifying the neural switch encoding. ParametricSlimmableContainer,HyperA2and older parametric NAM files, Full/Lite mode and static NAM export.- Input calibration, Auto output mode, corrected IR normalization and the independent Linear Volume attenuator.
- Session recall and the original JUCE VST3 / AU / Standalone plugin formats.
Native downloads
The release workflow builds and verifies seven native packages from the same commit, then publishes the GitHub release:
- Windows x64: VST3 and Standalone.
- macOS universal (Apple Silicon and Intel): VST3, AU and Standalone.
- Linux x86_64: VST3 and Standalone.
Windows/macOS binaries are not code-signed or notarized. See the README for installation instructions.
NAM Parametric Plugin 0.5.1
Single-network linear weight smoothing
- Improved noise rejection when moving knobs, HyperWaveNet/HyperA2 now use one neural audio model, not two copies with crossfade.
- The hypernetwork recalculates target weights at most once every 100 ms of model audio while controls move; intermediate knob changes are coalesced to the latest destination.
- Convolution coefficients interpolate linearly at every audio sample over 100 ms; the original trained weights and sound at stationary control positions remain unchanged.
- Removes the old Eco output-step correction, two-network Crossfade DSP, and experimental Control smoothing, timing and curve settings. The same fixed 100/100 ms Linear behavior is used regardless of older session settings.
- The independent measured Linear Volume output attenuator remains available, with its own fast gain smoothing and without regenerating neural weights.
Switch metadata and backwards compatibility
- New parametric models may specify each switch's
optionsas objects withname,value(original neural switch state) andindex_sw(zero-based visual order). For example{"name":"Crunch","value":1,"index_sw":0}. - Renaming/reordering a switch option in the display does not change its neural one-hot encoding, static-export position, or session value.
- Still loads older
enum_namesarrays andHyperA2/ParametricSlimmableContainerexports. Optionalconfig.params[].indexcontinues to determine knob/switch/Linear Volume presentation order. - Existing sessions continue to restore model/knob/IR/calibration settings; obsolete experimental transition settings are ignored.
Other included improvements
- Preserves the 0.5.0 IR unity-gain correction, output modes including Auto, input calibration, Full/Lite support, static NAM export and original metadata.
- No model retraining or conversion is required for old captures.
Native packages
- Windows x64: VST3 and Standalone
- macOS universal (Apple Silicon and Intel): VST3, AU, Standalone
- Linux x86_64: VST3 and Standalone
These are seven separately packaged native builds from the same verified release commit. Windows and macOS binaries are unsigned and are not notarized.
NAM Parametric Plugin 0.5.0
NAM Parametric Plugin 0.5.0
Version 0.5.0 updates parametric-model compatibility and corrects the large output-level drop that could occur when enabling an impulse response (IR).
What's new
New ParaNAM Trainer model format
- Loads the new
ParametricSlimmableContainerformat produced by current ParaNAM Trainer exports. - Preserves compatibility with legacy
HyperA2andParametricOutputGainmodels. - Supports optional
extensions.linear_volumecurves: the measured clean volume taper is applied after the neural model, with smoothing to avoid clicks. Volume-only adjustments do not require regenerating neural-network weights. - Keeps Full/Lite model selection, input calibration, session recall, and static NAM export.
IR loader: corrected output gain
- Compensates for the -18 dB attenuation previously introduced by AudioDSPTools' IR processing path.
- Uses the same IR normalization target as the Compare page in ParaNAM Trainer: average spectral power across 64 logarithmically spaced bands from 100 Hz to 5 kHz.
- Calculates the correction outside the real-time audio thread and updates it when the sample rate changes.
- Leaves the IR stage optional and keeps the plugin's output-gain control independent.
Note: The target is approximately 0 dB average spectral gain over the comparison band, not a guarantee of identical peak level or perceived loudness for every IR. An IR's frequency response can still change perceived volume.
Regression coverage and packaging
- Adds IR gain-normalization regression tests.
- Verifies model loading, audio processing, static export, and calibration/control behavior in native builds.
Downloads
Seven native packages are built and published only after their platform checks pass:
- Windows x64: VST3 and Standalone
- macOS universal (Apple Silicon + Intel): VST3, AU, and Standalone
- Linux x86_64: VST3 and Standalone
Windows and macOS binaries are not code-signed/notarized; see the README for installation instructions.
Full Changelog: v0.4.5...v0.5.0
Updated binaries — 10 November 2026
The seven downloadable binaries have been rebuilt from commit c2c426b14418 and re-verified on Windows, macOS (Intel + Apple Silicon) and Linux. These updated packages support the optional PNAM neural control index metadata, sharing one presentation order with Linear Volume; model processing, saved presets and static NAM export keep the original neural parameter ordering. Older PNAMs retain their historical control layout.
Source provenance: The original v0.5.0 Git tag and GitHub auto-generated source archives remain unchanged; the updated binary assets come specifically from the linked commit above.
SHA-256 of refreshed downloads:
NAM-Parametric-Plugin-0.5.0-Linux-x86_64-Standalone.tar.gz:9dd6c0caa09045265447ea37481c6e01985c8aaa5ec4edaf728d034919cbf14fNAM-Parametric-Plugin-0.5.0-Linux-x86_64-VST3.tar.gz:fd0827edf70042c5dc8df4fac2c41c195e1cb748fbbc471c51635d8cbe91673aNAM-Parametric-Plugin-0.5.0-Windows-Standalone.zip:3fd826140e88035d0fa06903175bb46238cffacf9d79c166fb998e1ee3eb666cNAM-Parametric-Plugin-0.5.0-Windows-VST3.zip:21b76e78fd7f675d98156ce36c437b31efc4c3b0133a44c8b6dabf669dd4d351NAM-Parametric-Plugin-0.5.0-macOS-AU.zip:f9b4922b0773be01f2d23b90c4ac6db515c6c4afe2588b6e505d8a67ef51299bNAM-Parametric-Plugin-0.5.0-macOS-Standalone.zip:541a3730869f3642617c5213f11f0ddd289f294e137343f255fd445dc3ed547eNAM-Parametric-Plugin-0.5.0-macOS-VST3.zip:795174d22b432de149a4a12718c6723a14b202952d25fbb60783c212d113ad0d
NAM Parametric Plugin 0.4.5
Preview 0.4.5 supports measured clean output-volume curves exported by ParaNAM Trainer 0.7.0. Volume is applied after the neural network with 5 ms smoothing; volume-only changes avoid neural-weight regeneration. Full/Lite, calibration, resampling, Eco transitions and static export remain supported. All seven packages passed native tests: Windows standalone/VST3, universal macOS standalone/VST3/AU, Linux x86_64 standalone/VST3. Models using ParametricOutputGain require this version or a compatible runtime.
NAM Parametric Plugin 0.4.4
HyperA2 playback now uses the specialized A2 audio backend for both Full and Lite,
with NeuralAudio-style multi-frame weight reuse for small x64 buffers and the
faster Eigen kernel for larger buffers. Updating knobs and switches preserves
convolution history without recreating the audio network.
- Reduced processing cost: a repeated synthetic-model benchmark on i7-12700H
measured 16–32% less CPU time for Full and 53–65% for Lite. Actual gains depend
on CPU, buffer size and hypernetwork size. - Preserves calibration, metadata, Full/Lite selection, Eco de-click, crossfade
mode, resampling and static export. No retraining or file conversion is needed. - Supports existing
.pnamand.namfiles; other architectures retain the
generic backend. No added block latency or AVX2 requirement. - Shares the optimized core and validated HyperA2 loader with the
NeuralAudio PNAM hardware fork.
Validation includes Full/Lite PyTorch golden audio, generic/optimized audio under
control automation, all 30 combinations of three independent switches mixed with
knobs, calibration, de-click/crossfade and static export.
Packages: Windows x64 standalone/VST3, macOS universal (Intel + Apple Silicon)
standalone/VST3/AU, and Linux x86_64 standalone/VST3.
NAM Parametric Plugin 0.4.3
Version 0.4.3 refreshes the interface with a #bf0000 red accent and neutral gray surfaces of similar brightness. The TTS logo from NAM On Steroids is the application icon. Includes the existing ASIO standalone backend, HyperA2 Full/Lite, .pnam and .nam support, calibration and Eco de-click. Audio behavior is unchanged. Windows x64 standalone and VST3; macOS universal standalone, VST3 and AU; Linux x86_64 standalone and VST3. Export, calibration and control-transition tests run on all platforms before publication.
NAM Parametric Plugin v0.4.1
Eco de-click is now the initial Control smoothing setting: one running audio network and an experimental 5 ms correction of control-induced output steps. Settings also offers the previous Crossfade mode for comparison (two running networks, higher CPU). Both .pnam parametric exports and legacy .nam files can be loaded, retaining all model metadata. The selection is recalled with the session; Full remains the initial HyperA2 size. Windows x64, macOS universal (arm64 + x86_64) and Linux x86_64 packages. Audio, calibration, export and control-transition tests run on all three platforms. Linux builds are tested on Ubuntu 22.04. See the README for installation instructions. Windows and macOS binaries are unsigned.
Windows packages refreshed in place with ASIO support, using the bundled JUCE ASIO backend. The build checks ASIO backend availability plus audio, calibration and export tests. Windows rebuild source: 6eff901. macOS and Linux packages are unchanged.