ComfyUI-Egregora-Audio-Super-Resolution

ABX Judge

Compares the supplied A/B guess with the identity recorded in abx_meta and returns the normalized identity, guess, and correctness in a DICT.

Egregora/Listening · 1 input · 1 parameter · 1 output

ABX Prepare

Trims both audio inputs to their shared length, clips them at the configured start and duration, chooses A or B as X using random_seed, and returns the clips and ABX metadata.

Egregora/Listening · 2 inputs · 3 parameters · 4 outputs

Audio Align (XCorr)

Resamples processed audio to the reference rate when needed, estimates delay using GCC-PHAT cross-correlation, applies delay compensation, and returns aligned audio, delay values, a peak_corr value of 0.0, and a debug plot.

Egregora/Analysis · 2 inputs · 3 parameters · 5 outputs

Audio Gain Match

Measures reference and input levels using integrated LUFS or RMS, applies a gain difference limited by max_gain_db, and returns matched audio with gain and level values.

Egregora/Analysis · 2 inputs · 1 parameter · 4 outputs

Audio Gain Match (1770)

Measures reference and input levels using integrated LUFS or RMS, applies a gain difference limited by max_gain_db, and returns matched audio with gain and level values.

Egregora/Analysis · 2 inputs · 2 parameters · 4 outputs

Audio Null Test

Optionally scales and inverts the processed audio, subtracts it from the reference, and returns the null audio with selected correlation, loudness, RMS, LSD, and residual metrics.

Egregora/Analysis · 2 inputs · 8 parameters · 2 outputs

Audio Plotter

Creates waveform, spectrogram, and difference-spectrum images from reference, processed, and null audio, with toggles for each plot.

Egregora/Visualization · 3 inputs · 5 parameters · 3 outputs

Egregora_DAC_Decode

Decodes DICT produced by Egregora_DAC_Encode back to AUDIO.

Egregora/Codecs · 1 input · 1 parameter · 2 outputs

Egregora_DAC_Encode

Encodes audio with DAC and returns latent 'z' & metadata in a DICT. Auto-downloads weights for chosen model_type on first use.

Egregora/Codecs · 1 input · 2 parameters · 2 outputs

Egregora_DeepFilterNet_Denoise

DeepFilterNet denoiser (speech enhancement) for ComfyUI. • Runs DeepFilterNet at 48 kHz (its native rate), using tensor I/O. • Mono or stereo (per-channel or downmix to mono before DFN). • Adds 'strength' wet/dry mix with equal-power or linear curve. • Adaptive mix driven by VAD (RNNoise if available, else energy/RMS proxy). • Post-gain (dB) and a simple peak ceiling limiter.

Egregora/Enhance · 1 input · 9 parameters · 1 output

Egregora_RNNoise_Denoise

RNNoise denoiser (speech-focused), ComfyUI node. • Runs at 48 kHz (10 ms = 480 samples). • Mono/stereo: per-channel or downmix to mono. • Uses pyrnnoise>=0.3.x 'denoise_chunk' API. • Adds static strength + adaptive mix (driven by per-frame VAD) + post-gain with ceiling.

Egregora/Enhance · 2 inputs · 7 parameters · 1 output

Egregora_WPE_Dereverb

Weighted Prediction Error dereverberation. Works mono or multi-channel. Uses STFT -> WPE -> iSTFT.

Egregora/Enhance · 1 input · 6 parameters · 1 output

EgregoraAudioUpscaler

FlashSR audio upscaler for musical content: resamples input to the model rate, processes it in chunks, and can resample the output to the selected sample rate.

1 input · 2 parameters · 1 output

EgregoraFatLlamaCPU

Spectral Enhance (Fat Llama — CPU/FFTW) — Pure CPU path using pyFFTW backend; no CUDA/CuPy required. — If you feed non-WAV inputs via path/URL, ffmpeg on PATH may be required by the package. :contentReference[oaicite:2]{index=2}

1 input · 6 parameters

EgregoraFatLlamaGPU

Spectral Enhance (Fat Llama — GPU only) - Normalize is always ON (clamps final amplitude and prevents clipping). - Adaptive filter disabled for speed (still available in library if you want a "slow" node).

1 input · 8 parameters

Loudness Meter (BS1770)

Returns a metrics DICT with integrated, momentary, and short-term LUFS, loudness range, and optionally true peak.

Egregora/Analysis · 1 input · 2 parameters · 1 output

Metrics (LSD + SI-SDR)

Computes enabled LSD mean and 95th-percentile values and/or SI-SDR from the mono means of reference and processed audio over their shared length.

Egregora/Analysis · 2 inputs · 4 parameters · 1 output

Null Test (Full)

Combines alignment, gain matching, null-test metrics, and plotting for reference and processed audio, returning aligned and matched audio, null audio, delay and gain values, metrics, and images.

Egregora/Analysis · 2 inputs · 12 parameters · 8 outputs

Resample Audio (HQ)

Resamples audio to target_sr using the selected mode; auto mode chooses an available polyphase or torchaudio method, with linear interpolation as a fallback, and returns unchanged audio when the rates match.

Egregora/Utils · 1 input · 3 parameters · 1 output