You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

SIREN-FX

Neural Audio Effects Processing with State Space Models

SIREN-FX is part of the SIREN Audio Suite - a family of neural audio processing models designed for professional music production workflows.

Model Description

SIREN-FX learns to model and apply complex audio effects using a State Space Model (S4) architecture. Unlike traditional neural audio effects that are limited to simple nonlinearities, SIREN-FX can capture time-dependent effects with theoretically unlimited receptive fields, making it ideal for:

Reverb - Room acoustics and spatial effects
Delay - Time-based echo and repetition
Chorus/Flanger/Phaser - Modulation effects
Complex effect chains - Multi-effect processing

Architecture

Component	Details
Base Architecture	S4 (Structured State Space Sequence Model)
Model Dimension	128
Number of Layers	8
State Size	64
Parameters	~1.5M
Sample Rate	44.1 kHz

The S4 architecture provides:

Unlimited receptive field - Can model any length-dependent effect
Linear time complexity - Efficient inference
Stable training - Proper initialization for long sequences

The SIREN Family

SIREN-FX is part of a suite of audio AI models:

Model	Purpose
SIREN-FX	Neural audio effects (this model)
SIREN-FIX	Audio restoration and repair
SIREN-MASTER	Audio enhancement and mastering
SIREN-STEER	Steerable audio transformations
SIREN-SEPARATE	Source separation
SIREN-TRANSCRIBE	Music analysis (key, tempo, beats)

Usage

import torch
import torchaudio

# Load model
checkpoint = torch.load('siren_fx.pt', map_location='cpu')
model_state = checkpoint['model_state_dict']

# Model expects mono audio at 44.1kHz
# Input shape: (batch, 1, samples)
# Output shape: (batch, 1, samples)

Training Details

Training Data: Large-scale audio effects dataset
Training Duration: 200 epochs
Hardware: NVIDIA B200 GPUs
Final Validation Loss: 1.0482

Intended Use

SIREN-FX is designed for:

Music production and post-production
Audio effect modeling and emulation
Creative sound design
Research in neural audio processing

Limitations

Optimized for 44.1kHz sample rate
Best results with mono input
Effects learned from training distribution

License

Apache 2.0

Citation

If you use SIREN-FX in your research, please cite:

@software{siren_fx_2026,
  title={SIREN-FX: Neural Audio Effects with State Space Models},
  author={SIREN Team},
  year={2026},
  url={https://huggingface.co/hilarl/siren-fx}
}

Contact

For questions and feedback, please open an issue on the model repository.

Downloads last month: -; Downloads are not tracked for this model. How to track