It seems that version 3.10 of the SWAM string instruments has introduced a new feature — IR (Impulse Response).
I’d like to jot down what I’ve found out so far as Part 2 of this topic.
If you notice anything inaccurate, please feel free to point it out.
1. SWAM design and conventional mechanism (~v3.9)
SWAM is a fully physically modeled instrument that reproduces the structure and playing behavior of real acoustic instruments in real time.
Based on inputs such as bow movement, pressure, and finger position, it continuously calculates string vibration, body resonance, and overtone variations.
As a result, it produces slightly different sounds every time—hard to quantify, but undeniably more organic.
While CPU usage is heavy, the library size remains impressively small.
In earlier versions, SWAM included Ambience, an internal room simulator.
It allowed users to switch acoustic spaces with city-named presets like Cremona, Firenze, or Roma.
However, these did not rely on IRs; the reverberation was entirely generated by algorithmic, numeric modeling.
When Ambience was turned off, spatial reverb disappeared, but the physical modeling of the instrument itself remained intact—external IRs could then be added later within a DAW if desired.
Version 3.8 (2024) introduced an updated Ambience Room Simulator, enabling more sophisticated room simulations, but still without using IRs.
2. The Introduction of Instrument Body IR in SWAM v3.10 (September 2025)
Finally, in version 3.10 (September 2025), Instrument Body (IR) has been added.
This simulates the impulse response of the instrument’s body, aiming to recreate the resonance of the wooden box itself more faithfully.
New presets—Cremona, Firenze, Pisa, Bologna, and others—have been introduced, focusing on the character of the instrument’s body, distinct from room ambience.
Each instrument now also includes a Dry Body option, allowing users to bypass the IR entirely and keep the older “dry” modeling behavior. The internal signal flow is now structured as: Physical Modeling → Instrument Body (IR) → Ambience → Output. This cleanly separates the two layers: the body resonance (IR) and the room/hall reverberation (Ambience).
This addition makes bow movement and attack nuances sound more natural—particularly in solo performances, it promises a noticeably more “acoustic” realism.
A Brief Look Back: Sample Modeling and Audio Modeling
1. The Birth of Sample Modeling
Around 2007, a group of Italian engineers founded Sample Modeling.
They pioneered an innovative hybrid approach—using very short audio samples, then shaping and correcting them through physical modeling.
Their brass and woodwind instruments, such as The Trumpet and The Saxophone, drew wide attention for combining sample realism with modeling flexibility.
2. The Founding of Audio Modeling
Around 2017, several of Sample Modeling’s co-founders branched off to form Audio Modeling, where the current SWAM engine was born.
This new team pursued a completely sample-free physical modeling philosophy.
Thus emerged two distinct paths:
• Sample Modeling → hybrid of samples + modeling
• Audio Modeling (SWAM) → pure physical modeling
3. The Different Approaches to IR
This historical split can also be seen in how each company treats IRs.
This historical divergence is also evident in the way IR is handled。
Sample Modeling (SCES)
→ 。 gives "spatial characteristics" by applying an internal IR (convolution reverb) to a sample recorded in an anechoic chamber.
→ 。 with CC100 to control IR contribution
Audio Modeling (SWAM)
→ Initially no IR, only Ambience (numerical model room simulation) 。
→ In v3.10 in 2025, Instrument Body (IR) will be introduced for the first time, evolving into a two-tiered structure separated from Ambience (space)。
It is interesting to note that although the origin is the same, we arrived at the difference between SCES = IR of space and SWAM = IR of instruments。
Some time ago, I performed The Swan by Saint-Saëns using SWAM Solo Cello.
Now that this new IR feature has arrived, I’m tempted to revisit that piece and compare the results.
Of course, I’m still very much an amateur when it comes to sequencing—so we’ll see how it turns out!