ComfyUI Extension: ComfyUI · Egregora Audio Super‑Resolution

Authored by lucasgattas

Created

Updated

36 stars

High‑quality music audio enhancement for ComfyUI: FlashSR super‑resolution + Fat Llama spectral enhancement (GPU & CPU).

Custom Nodes (0)

    README

    🎧 ComfyUI — Egregora Audio Super‑Resolution

    Bring music up to studio‑grade sample rates right inside ComfyUI.

    This repo ships three production‑oriented upscaling/enhancement nodes and bundles a set of integrated utility toolsets (enhance, evaluation, null‑testing) so you can denoise → upscale → measure without wiring a huge graph.

    ✨ What’s inside

    custom_nodes/
  ComfyUI-Egregora-Audio-Super-Resolution/
    __init__.py
    egregora_audio_super_resolution.py   # FlashSR node
    egregora_fat_llama_gpu.py            # Fat Llama (CUDA/CuPy)
    egregora_fat_llama_cpu.py            # Fat Llama (CPU/FFTW)
    egregora_audio_enhance_extras.py     # RNNoise / DeepFilterNet / WPE / DAC
    egregora_audio_eval_pack.py          # ABX, Loudness/Match, Metrics, HQ Resample
    egregora_null_test_suite.py          # Align, Gain‑Match, Null, Plots
    flashsr_min.py                       # Light wrapper for FlashSR
    install.py                           # Repo + weights/deps bootstrapper
    requirements.txt
    deps/
      FlashSR_Inference/                 # pulled automatically on install

    Core nodes

    • Audio Super Resolution (FlashSR) — one‑step diffusion upsampler (music‑friendly) ⚡
    • Spectral Enhance (Fat Llama — GPU) — CUDA/CuPy accelerated iterative spectral enhancer 🐍🧪
    • Spectral Enhance (Fat Llama — CPU/FFTW) — portable CPU fallback using pyFFTW 🧠

    Integrated utility toolsets (used inside the SR nodes)

    • Enhance — Extras

      • RNNoise Denoise (48 kHz, adaptive mix, strength, post‑gain)
      • DeepFilterNet 2/3 Denoise (48 kHz native)
      • WPE Dereverb (nara‑wpe)
      • DAC Encode/Decode (Descript Audio Codec)

    • Eval Pack

      • ABX prepare/judge clips
      • Loudness meter (BS.1770), Gain‑Match (LUFS/RMS)
      • Metrics: SI‑SDR, Log‑Spectral Distance (LSD)
      • High‑quality resampler (SciPy/torch fallbacks)

    • Null Test Suite

      • Align (XCorr GCC‑PHAT), Gain‑Match, Null, difference plots

    These helpers are wired so you can ABX / null‑test right from the SR node panel.

    🧩 Install (ComfyUI portable or venv)

    1. Copy the folder to ComfyUI/custom_nodes/ and restart ComfyUI once.

    2. Install Python deps using ComfyUI’s Python:

    # From ComfyUI root
python -m pip install -r custom_nodes/ComfyUI-Egregora-Audio-Super-Resolution/requirements.txt
python custom_nodes/ComfyUI-Egregora-Audio-Super-Resolution/install.py

    • We do not install torch/torchaudio here to avoid breaking ComfyUI’s CUDA build.

    • First run will:

      • clone deps/FlashSR_Inference/
      • check for FlashSR weights
      • warm up DeepFilterNet / DAC / RNNoise caches for smoother first use
    1. FlashSR repo & weights

    • The node pulls the upstream inference code automatically into deps/FlashSR_Inference/.

    • This node does not include FlashSR code or weights. The commonly referenced FlashSR_Inference repo currently lacks a license. Unless you have explicit permission from the rights holder(s), do not use FlashSR code/weights for commercial purposes. Proceed at your own risk.

    • Place weights in ComfyUI/models/audio/flashsr/ with exact filenames:

      • student_ldm.pth, sr_vocoder.pth, vae.pth

    • Or set an env var to auto‑download from your HF repo:

    # point to a HF repo containing those three files
# Windows (cmd)
set EGREGORA_FLASHSR_HF_REPO=yourname/flashsr-weights
# macOS/Linux
export EGREGORA_FLASHSR_HF_REPO=yourname/flashsr-weights
    1. GPU extras (for the Fat‑Llama GPU node)

    Install a CuPy wheel matching your CUDA (example for CUDA 12):

    python -m pip install "cupy-cuda12x>=13.0"

    If Windows shows NVRTC / vector_types.h errors, install the CUDA runtime DLL wheels:

    python -m pip install -U nvidia-cuda-runtime-cu12 nvidia-cuda-nvrtc-cu12 \
  nvidia-cublas-cu12 nvidia-cufft-cu12 nvidia-curand-cu12 \
  nvidia-cusolver-cu12 nvidia-cusparse-cu12
    1. FFmpeg

    Ensure FFmpeg is on your PATH for reading/encoding audio.

    📦 Requirements

    requirements.txt keeps things lean:

    • Core: soundfile, numpy, tqdm, requests, huggingface_hub
    • SR/enhance: fat-llama, fat-llama-fftw, pyrnnoise, deepfilternet (import as df), nara-wpe (import as nara_wpe), descript-audio-codec
    • Optional: scipy for HQ resampler/metrics

    Booleans in node UIs use the BOOLEAN datatype in INPUT_TYPES (proper toggle).

    🛠️ Nodes & key settings

    1) Audio Super Resolution (FlashSR)

    • Chunks → overlap‑add → stitches to 48 kHz (or chosen target).
    • Inputs: chunk_seconds (default 5.12), overlap_seconds (0.5–0.75 if seams), device, target_sr, output_format, audio_path / audio_url, flashsr_lowpass (gentle LPF).
    • Outputs: AUDIO buffer + saved file.

    2) Spectral Enhance (Fat Llama — GPU/CPU)

    • Iterative soft‑thresholding with spectral post.
    • Inputs: max_iterations, threshold_value, target_bitrate_kbps, toggle_autoscale, target_format, audio_path / audio_url.
    • Outputs: AUDIO buffer + saved file.

    Utility toolsets (used inside SR nodes)

    • Denoise/Dereverb: RNNoise, DeepFilterNet 2/3, WPE
    • Codec: DAC encode/decode
    • Eval: ABX clips + judge, BS.1770 loudness, gain‑match, SI‑SDR, LSD
    • Null: Align → match → null + difference plots

    🎚️ Quality tips (music)

    • FlashSR first, Llama second: upscale to 48k, then a light Llama pass (iterations≈200, threshold≈0.5) if you want a touch of sparkle.
    • Overlap: If you hear ticks between chunks, raise overlap_seconds a bit.
    • Don’t over‑iterate: very high iterations/threshold can sound brittle.

    🔍 Licenses (upstream projects)

    • Fat‑Llama / fat‑llama‑fftw: BSD‑3‑Clause (see PyPI).
    • FlashSR_Inference: check upstream repo for license status.
    • This ComfyUI integration is licensed as per this repository’s LICENSE.

    🧪 Troubleshooting

    • FlashSR import error: delete deps/FlashSR_Inference/ and restart to re‑bootstrap.
    • Missing FlashSR weights: place the 3 files in models/audio/flashsr/ or set EGREGORA_FLASHSR_HF_REPO.
    • CUDA/CuPy NVRTC errors (Windows): install the nvidia-*-cu12 runtime wheels listed above and ensure your CuPy wheel matches CUDA.
    • FFmpeg not found: install FFmpeg and ensure it’s on PATH.

    🙌 Credits

    • FlashSR research & inference code by the original authors.
    • Fat Llama packages by RaAd (PyPI maintainer).
    • ComfyUI integration & node UX by Egregora.

    Happy upsampling! 🎶

    📜 Changelog

    • v0.2.0 — Added Enhance/Eval/Null toolsets; new installer + warmups.
    • v0.1.0 — Initial release: FlashSR SR node, Fat Llama GPU/CPU.