ADSRPre-EQ / Tone Stack
Envelope roles become fixed spectral contours, rather than time-varying envelope generators.
Reinterpreting Synthesizer Parameter Spaces for Expressive Tone-Shaping Audio FX Design
CTAG-FX grew out of my interest in exploring and shaping guitar tones through audio effects.
Each configuration defines a reusable tone-shaping processor for new audio inputs.
Geonung Jo · Jongeun Choi
CTAG-FX explores a different approach to text-guided audio effects: constructing the control space of an FX processor rather than adjusting parameters within a predefined FX chain.
REPRESENTATIVE SYNTH-TO-FX REINTERPRETATIONS
Envelope roles become fixed spectral contours, rather than time-varying envelope generators.
Contrasting source roles become smooth and aggressive nonlinear coloration; no oscillator audio is transferred.
Influence relationships become bounded offsets across drive, EQ and noise controls.
The mixer combines a clean path, a processed path, and an independent noise path before the final output stages.
Synthesizer parameter roles become controls for EQ, drive, saturation, mixing, and input/output gain.
External audio supplies the signal; oscillator output is not transferred.
The Mod Matrix links four static character axes to five targets through bounded control offsets.
It shapes control relationships, rather than adding time-varying modulation to the audio path.
Explore how sonic character can be reinterpreted as FX behavior.
Compare a reference, its A&R-CTAG synth representation, and the resulting guitar FX.
Role-based mapping produced prompt-specific spectral and nonlinear behavior; scrambling reduced separation.
The processor test bypassed A&R-CTAG: GPT-4o directly instantiated configurations from four tone prompts. Ten configurations per prompt were evaluated on a held-out 10-second guitar recording.
The ablation retained parameter values and permuted their control assignments, using 30 shared permutations.
Ten prompts, equal 300-iteration budgets. These LAION-CLAP results concern synth-side generation, not processed-guitar perceptual quality.
| Metric | CTAG | A&R-CTAG | Change |
|---|---|---|---|
| Initial score mean | 0.1449 | 0.2981 | +105.7% |
| Final score mean | 0.6166 | 0.7545 | +22.4% |
| Final score SD | 0.1246 | 0.0595 | −52.2% |
| Score change mean | 0.4718 | 0.4564 | −3.3% |
Signal-level analysis and embedding similarity are reported here; perceptual correspondence remains to be evaluated. See the preprint.
01 / CREATOR
02 / ENGINEER
03 / MUSICIAN
EXPLORE THE FULL RESEARCH