Elforyn2 - Piano VST DEMO

Elforyn2 is a physically modelled acoustic piano. There are no samples anywhere in it. Each of the 88 keys is a permanent object: its own unison strings as a bank of modal resonators, its own damper, its own place on the bridge. A note happens because a time-domain nonlinear felt hammer collides with those strings and the two are solved against each other through contact — so velocity changes the tone and not merely the volume, restrikes interact with a string that is already moving, and the sustain pedal lets every other string in the instrument ring in sympathy because they were always there.

What that buys, briefly:

• Real felt. A hardening, hysteretic hammer whose mass, stiffness, strike point and contact time are graded across the compass — heavy soft bass hammers, hard light treble. Contact time is around 4 ms in the bass and under 1 ms at the top, which is where the dynamic timbre comes from.
• Real strings. Inharmonicity, two decay rates, unison detune and the resulting shimmer/re-bloom, duplex (aliquot) resonance driven by the note's own live string sum, combination-tone "metallic bite" in the bass, and a two-polarisation double decay — the prompt sound falls away and a quieter, slightly detuned aftersound remains under it.
• A real board. The strings terminate on a modal soundboard with an actual mechanical admittance, so the board absorbs energy, colours the tone and couples strings to each other. Bending-wave attack dispersion (treble radiates first, lows trail) falls out of the board's own modes rather than being drawn on afterwards.
• A real room, captured. Up to five placeable virtual microphones — position, height, distance, polar pattern, level, pan, room send. They sit in one shared modelled room, so a close spot mic reads dry and a distant one reads roomy, from geometry. Key thumps, damper noise and pedal noise are sources in that room too, imaged through whichever mics you have up. No mics enabled means silence, as it would with real microphones.
• Continuous (half) sustain pedal and una corda; a grand model plus a first upright measured from real recordings.

Everything that could be measured against real recordings was. Where the engine disagreed with the reference, the reference generally won. Inharmonicity and decay times are inside the measured bands across the whole keyboard, and the dynamic range from fff to mf now sits within a decibel of the real instrument at every note. Several ideas that seemed good were built, measured, sounded wrong, and were removed — the code carries a few of those gravestones on purpose.

Where it falls short of the best in class (i.e. Pianoteq)

I'd rather you hear this from me than find it out after paying:

• One and a half instruments. A grand, plus an upright that is still being voiced. Pianoteq gives you dozens of instruments including historical ones. That gap is enormous and won't close soon.
• Windows VST3 only. No macOS, no AU, no Standalone, no Linux. Developed and tested primarily in Reaper.
• The GUI is functional, not beautiful. Tabbed panels of knobs. No graphical mic placement, no per-note editing curves, no keyboard graphics. It will get a design pass; it hasn't had one.
• CPU. It runs comfortably in real time with margin, but it is not in Pianoteq's league for optimisation. Pedal-down glissandi and dense fortissimo bass clusters are where you'd notice.
• Missing features you may expect: no microtuning/Scala, no MPE, no sostenuto pedal, no multichannel output bus, no built-in effects beyond the board/mic/room chain, and a small preset bank that is mostly different capture setups rather than different pianos.
• Known tonal residuals, measured - still have a way to go to match some of the audio signatures of reference audio I’m using to compare against, but making good progress.
• In short I don’t think I can compete with those guys - they have PhDs with years of experience, and I have a good ear and passable coding skills. At best. This is for people who don’t need the best, but need something that’s hopefully passable.