In this article, I explore two leading physically modeled string libraries: SWAM Strings by Audio Modeling and Solo, Chamber & Ensemble Strings (SCES) by Sample Modeling.
While working on MIDI sequencing of Beethoven’s piano sonatas, I’ve also spent time experimenting with string sequencing—such as violin and cello concertos, and string quartets. Despite trying various libraries, I still haven’t found one that feels like the definitive solution. Of course, that may be because I haven’t fully mastered any of them yet.
This time, I decided to focus specifically on physically modeled string libraries, aiming to deepen my understanding from the perspective of IR (Impulse Response). I’m not a professional, so please feel free to correct any inaccuracies you may find.
Libraries Under Review:
- Audio Modeling: SWAM Solo Strings, SWAM String Sections
- Sample Modeling: Solo, Chamber Ensemble Strings (SCES)
Evaluation Focus
Both libraries excel at replicating detailed articulations and playing techniques. However, in the end, what matters most to me is how convincingly the overall sound resembles a real acoustic instrument.
In my experience, SWAM tends to shine in fast, energetic pieces with strong attacks. On the other hand, SCES often sounds more natural in slower, more static music. In this post, I’ll explore some of the structural reasons behind these differences.
1. Audio Modelng - SWAM
Concept and Approach:
SWAM is a fully physical modeling instrument. It simulates the acoustic behavior of real instruments in real time by mathematically modeling their physical structures and performance gestures.
Key Features:
Modeling of energy sources (Exiter)
- Bow speed, pressure, position, and direction
- Pizzicato finger placement, attack, and force.
Resonator modeling
- Resonance of the instrument body, f-holes, air column, etc.
- Includes real-time calculations of standing waves and resonant frequencies.
Interaction
- Friction between bow and string, noise generation, and other nuanced behaviors
Dynamic Response
- Not just changes in volume, but also in tonal color and harmonic content
- Strong bowing produces a rougher tone, while gentle playing results in delicacy
Because all of these parameters are computed in real time using mathematical formulas, each performance produces a slightly different sound, lending SWAM a vivid, “live” character. However, this also makes it CPU-intensive. Fortunately, the library itself is extremely lightweight - just a few dozen megabytes.
About SWAM's acoustic space and IR:
In SWAM, there are options that look like "instrument bodies" such as "CREMONA", "FIRENZE", and "ROMA", which are actually "Ambiente", a room simulation function unique to SWAM that changes the acoustic space. This is actually a unique room simulation feature of SWAM called "Ambiente" that changes the acoustic space. This is not IR-based, but a real-time calculation based on a physical model.
In other words, when Ambiente is turned off, the spatial resonance is removed and the instrument is left in a dry state. In this case, it is possible to incorporate an IR loader in the DAW and apply an external IR to further change the character of the sound.
2. SCES by Sample Modeling
Concept and Approach:
SCES is a hybrid engine that combines short recorded samples with real-time physical transformations. It aims to balance the authenticity of sampling with the flexibility of modeling.
Key Features:
Sampling
- Recordings of actual performances in very short units (several hundred ms)
- Used as basic material without covering a large number of techniques
- The work is performed in an anechoic chamber (anechoic chamber)
Real-time processing
- Pitch conversion: Pitch shift with formant correction
- Dynamics change: not only volume but also timbre and overtones change
- Transitions (legato, glissando, etc.)
- Changes in microphone position and radiation model
Advantages of the Hybrid Approach:
- More flexible performance expression than sample-based systems
- Lower CPU load and easier to handle than full modeling
- Relatively lightweight in terms of library size
Relationship with IR:
In SCES, impulse responses (IRs) are used to shape both spatial characteristics and timbral changes. These are controlled via CC100, which adjusts the amount and type of the internal IR. Since the instruments are recorded in an anechoic chamber, it is possible to treat the performance component and the spatial component to a certain extent as separate elements.
I previously assumed that setting CC100 to “0” would result in a completely dry signal with no IR applied. However, after consulting the developer’s support, I learned that both technically and musically, the minimum value of CC100 is intentionally limited to “1.” This design choice avoids a fully anechoic state and preserves a minimal amount of resonance necessary for the instrument to sound natural and musically viable.
Even at this minimum setting, a very slight room character remains, but in practical terms the sound is sufficiently dry. From this point, spatial characteristics can still be shaped externally by using an IR loader or reverb within the DAW, allowing the overall acoustic space to be customized and unified as needed.
Sample Modeling was established by splitting off from Audio Modeling, and their sound creation philosophies and methods are very different. It will be very interesting to see how both companies adapt to the advancement of AI technology in the future.
Note (January 25, 2026):
After further testing and confirmation with the SCES support team, I learned that CC100 is not designed to reach a true zero state. The minimum value of 1 is an intentional part of the instrument’s design. This article has been updated accordingly to reflect that understanding.