Dawesome Novum v1.21 [WiN]

• Product: Novum
• Developer: Dawesome
• Version: 1.21
• Format: VST3, AAX
• Requirements: Windows 10 or later
• Source: dawesomemusic.com/plugins/novum

Novum is a granular synthesizer whose core operation is Non-negative Matrix Factorization applied to audio spectrograms: it decomposes a loaded sample into six independent layers by distributing spectral frequency content across components that, when summed, reconstruct the original signal with full fidelity. Each layer carries its own extracted timbre and temporal envelope, both editable independently or linked across all six simultaneously. Cross-synthesis via drag-and-drop assigns the timbre content of one sample to the envelope structure of another. The Timbre Flower modifies spectral character per layer; HOMOGENIZE removes phase artifacts from edited layers; SYNTIFY reshapes harmonic structure toward a synthetic character. The grain engine operates with full control over density, position, jitter, blur, and size, with DAW-tempo sync for grain emitter rate. Six effects slots run in series or parallel, including Shimmer and Cloud reverbs, with full MPE and a comprehensive modulation network. It answers the query: how do I build evolving pads, cinematic drones, and hybrid textures from single samples without using a conventional granular scanner or sampler.

Key Takeaway

Composers working in cinematic, ambient, and hybrid contexts — building evolving beds, slow-motion textures, and tonal drones from short source material — are the functional match for Novum’s NMF decomposition workflow. The six-layer architecture with independent timbre and envelope per layer produces movement and tonal variation that a conventional single-voice granular processor doesn’t replicate by changing grain size or scan speed alone. The granular engine has an inherent diffuse character at wide emitter spreads that becomes a production constraint on sessions requiring tight transient definition or acoustic instrument reproduction; producers whose workflow centers on percussive rhythmic content or acoustic realism will reach that boundary faster than those building sustained textural material.

NMF Decomposition: What the Six Layers Actually Are

Standard granular synthesis takes a sample as-is and reads it in grains — the source material determines what the grains contain, and sonic character is shaped through grain playback parameters. Novum’s NMF stage happens before granular playback begins: the algorithm treats the sample’s spectrogram as a matrix of frequency-versus-time data and factorizes it into six component matrices whose product approximates the original.

Each component captures a different cross-section of the spectral content — timbral characteristics that tend to co-occur in the original material cluster together, producing layers that represent distinguishable sonic elements rather than arbitrary frequency bands or time segments. On a simple sustained tone, the layers resemble each other closely; on a complex source — a drum loop, a choir recording, a layered pad — they diverge, each retaining a recognizable strand of the original’s spectral behavior.

The six layers sum back to the original signal when all envelopes and timbres are left untouched; any modification diverges from that reconstruction, which is the mechanism that makes cross-synthesis and timbre exchange produce something other than noise. Each layer can be turned on, off, or solo’d independently before the grain engine processes it, letting the engineer isolate which spectral strand is contributing what to the final sound.

Timbre Flower, HOMOGENIZE, and the Spectral Editing Axis

Each layer has its own Timbre Flower — a circular controller with the extracted timbre at center, where moving upward softens spectral content, moving right increases noise-like character, and moving down reshapes the layer toward a conventional synthesizer-like harmonic structure. These movements are continuous and additive, so a layer can be pushed toward both softer and more noise-like simultaneously by positioning the controller between those axes.

HOMOGENIZE removes phase information accumulated by timbre editing or cross-synthesis, cleaning up layers that have developed discontinuities or gritty artifacts from multiple edit passes. Without it, heavy timbre manipulation on a complex source layer can introduce incoherence between the layer’s spectral content and its phase behavior, audible as roughness or tonal inconsistency.

Linking edits across all six layers simultaneously via the orange bracket controls applies Flower position, grain settings, and other per-layer parameters to the entire layer stack. This is useful for global spectral character shifts but overwrites individual layer settings that were configured independently before linking was enabled.

Cross-Synthesis: Envelopes From One Source, Timbres From Another

Dropping a second sample onto the timbre power buttons transfers its NMF-extracted timbral content onto the currently loaded sound’s envelope structure — a flute’s spectral character playing through the temporal dynamics of a piano, because the envelopes controlling energy movement come from the piano’s decomposition, not the flute’s. Dragging onto a single layer rather than all six replaces only that layer’s timbre, leaving the remaining five intact for progressive hybridization.

The Lock icons for timbre and envelope operate independently. Locking timbre while cycling through presets keeps the loaded sample’s spectral character constant while the preset’s synthesis parameters change around it; locking envelope keeps the temporal dynamics while the timbre and synthesis chain vary.

Both locks together produce a fixed spectral and temporal skeleton that preset browsing dresses with different synthesis configurations. This makes the Locks the fastest path to extracting dozens of usable variations from a single source sample without re-importing or re-decomposing.

Grain Engine Controls and DAW Sync

The grain engine controls start position, loop points, density, positional jitter, blur, and grain size per layer — all modulatable from the internal network. Density governs how many grains play simultaneously; at lower density, individual grain activations become audible as discrete amplitude events whose rhythm depends on grain size.

Grain emitters can sync to DAW tempo, which aligns activation timing to beat subdivisions and converts the rhythmic pulsing from a free-running artifact into a deliberate rhythmic structure. Speed and pitch are independent, so a drum loop can be transposed without changing tempo.

Positional jitter distributes grain start points randomly around a set position, producing texture from the spectral variation across different parts of the sample. Reducing jitter to zero locks grain playback to a narrow section, tightening the texture toward a single spectral character at the cost of movement.

SYNTIFY, Dirt, and the Syn Section

SYNTIFY manipulates a layer’s harmonic structure toward a synthesizer-like character by reducing spectral irregularities and inharmonic content produced by NMF decomposition of complex sources. Two character modes select between different regularization approaches; the more appropriate one is source-dependent rather than a fixed setting.

On choir or voice samples, SYNTIFY removes formant complexity and breath content; on field recordings or noise-adjacent sources, it introduces harmonic center into material with no original pitch. Dirt adds harmonic saturation at the Syn stage independently, covering the range from subtle enrichment to heavy clipping.

The analog resonant filter applies to the Syn section output before the FX stage. The six effects slots — Shimmer and Cloud reverbs in particular — add substantial spatial diffusion and can dominate the final patch character to the point where disabling them changes what the granular and NMF processing is actually contributing.

Where NMF Decomposition Stops Working

NMF decomposition produces consistent results on sustained, harmonically coherent source material. On transient-heavy or dynamically complex sources, the factorization distributes spectral energy in ways that can produce cross-layer phase artifacts, dense granular texture that resists clean synthesis, or decompositions where individual layers don’t carry identifiable sonic characteristics in isolation.

The granular character is present throughout — at wide emitter settings with high jitter, the output has diffuse, continuously evolving texture that suits ambient and cinematic contexts but works against percussive clarity or realistic acoustic instrument reproduction. Close-miked acoustic instruments with sharp attack transients produce results that are recognizably granular rather than recognizably instrumental.

CPU load is optimized but not negligible; on older hardware, multiple instances in a dense session may require bouncing to audio. The six effects slots don’t include convolution reverb, multiband compression, or transient processing — the effects chain is designed to augment granular texture, not replace a full channel processing workflow.

FAQs