OLDSKOOL978/universr-audio

UniverSR - General Audio (Native SafeTensors)

Vocoder-free audio super-resolution model that upsamples 8/12/16/24 kHz → 48 kHz audio using flow matching in the complex STFT domain. Trained on speech, music, and sound effects.

This repository provides verified, zero-overhead SafeTensors weights (model.safetensors) converted from the original checkpoint published by Choi et al., eliminating unsafe PyTorch pickle deserialization (pytorch_model.bin) while enabling instant memory mapping. Integrated and optimized for general Audio Super-Resolution & Restoration in Furgie V2.

• Paper: arXiv:2510.00771
• Original Code: github.com/woongzip1/UniverSR
• Implementation: github.com/oldskool978/Furgie

---

Optimal Inference Defaults

For optimal broadband music and general audio reconstruction:

• Target Delivery Mode: 48.0 kHz (32-bit Float PCM)
• Flow ODE Solver: 2nd-Order Midpoint Runge-Kutta (Midpoint RK2)
• Integration Steps: 16 steps
• Guidance Scale ($w$): 0.00 (Single-pass deterministic conditional trajectory)
• Conditioning Anchor: 24 kHz Anchor (Harmonic Inpainting: 12.0 kHz – 24.0 kHz)
• Headroom Strategy: Bypass (Passband Bit-Exact Unity 1.0x)

---

Usage

Via Furgie Production Harness

Clone the optimized Furgie repository:

git clone https://github.com/oldskool978/Furgie.git
cd Furgie
pip install -r requirements.txt

# Download and verify SafeTensors weights
python scripts/hydrate_models.py

# Launch interactive super-resolution harness
python harness.py

Python API Integration

import torch
from furgie_core.engine import FurgieEngine
from furgie_core.schema import FurgieRequest

engine = FurgieEngine(
    device="cuda" if torch.cuda.is_available() else "cpu",
    model_repo_id="OLDSKOOL978/universr-audio"
)

request = FurgieRequest(
    input_path="input_audio.wav",
    output_path="output_restored_48k.wav",
    ode_steps=16,
    solver="midpoint",
    guidance_scale=0.0,
    input_sr_anchor=24000,
    headroom_mode="bypass"
)

telemetry = engine.synthesize_request(request)
print(f"Restored audio generated at: {telemetry.output_path} (RTF: {telemetry.real_time_factor}x)")

---

Architectural Configuration

audio:
  target_sample_rate: 48000
  n_fft: 1024
  hop_length: 512
  win_length: 1024
  power_alpha: 0.2

model:
  dims: [96, 192, 384, 768]
  depths: [2, 2, 4, 2]
  time_dim: 256
  cond_dim: 384
  total_freq_bins: 512
  hr_freq_bins: 432
  feature_enc_layers: 4
  sr_to_lr_bins:
    8: 80
    12: 128
    16: 170
    24: 256

universr_flow_core:
  enabled: true
  repo_id: OLDSKOOL978/universr-audio
  solver: midpoint
  ode_steps: 16
  guidance_scale: 0.0
  input_sr_anchor: 24000

---

Citations

Original UniverSR Architecture

@inproceedings{choi2026universr,
  title     = {{UniverSR}: Unified and Versatile Audio Super-Resolution via Vocoder-Free Flow Matching},
  author    = {Choi, Woongjib and Lee, Sangmin and Lim, Hyungseob and Kang, Hong-Goo},
  booktitle = {IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP)},
  year      = {2026}
}

Furgie SafeTensors & Super-Resolution Harness

@software{furgie2026,
  author    = {OLDSKOOL978},
  title     = {Furgie: Optimal Transport Flow-Matching Audio Super-Resolution Harness},
  url       = {https://github.com/oldskool978/Furgie},
  year      = {2026}
}