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-TRANSCRIBE

Music Analysis with Self-Supervised Foundation Model

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

Model Description

SIREN-TRANSCRIBE analyzes music to extract musical key and tempo using a 330M-parameter self-supervised music encoder as a foundation model with custom classification heads.

Key capabilities:

Key detection - Identify musical key (24 classes: major and minor)
Tempo estimation - Accurate BPM detection
Foundation model - Built on 330M-parameter music encoder

Architecture

Component	Details
Base Model	330M-parameter music encoder (24 transformer layers)
Key Head	Custom MLP (24 classes)
Tempo Head	Custom MLP (regression)
Sample Rate	24 kHz

The SIREN Family

Model	Purpose
SIREN-FX	Neural audio effects
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 (this model)

Usage

import torch

# Load model
checkpoint = torch.load('siren_transcribe.pt', map_location='cpu')

# Model expects audio at 24kHz
# Output: {"key": "Am", "tempo": 120.5}

Training Details

Training Data: Large-scale music dataset with analysis labels
Hardware: NVIDIA B200 GPU
Training Duration: 100 epochs

Intended Use

Musical key detection
Tempo/BPM estimation
Music information retrieval
DJ tools and music organization
Research in music understanding

Limitations

Key detection limited to 24 major/minor keys
Tempo estimation best for 60-200 BPM range
Requires 24kHz input

License

Apache 2.0

Citation

@software{siren_transcribe_2026,
  title={SIREN-TRANSCRIBE: Music Analysis with Foundation Model},
  author={SIREN Team},
  year={2026},
  url={https://huggingface.co/hilarl/siren-transcribe}
}

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