Audio

274 public types in this namespace, organized by kind.

Models & Types (141)

Type Summary
ACEStep<T> ACE-Step accelerated consistency-enhanced music generation model.
ASTModel<T> AST (Audio Spectrogram Transformer, Gong et al.
AST<T> AST (Audio Spectrogram Transformer) model for audio event detection and classification.
AudioEvent Represents a detected audio event.
AudioEventDetector<T> Audio event detection model for identifying sounds in audio (AudioSet-style).
AudioFlamingo2<T> Audio Flamingo 2 multimodal audio-language model for audio understanding with interleaved inputs.
AudioGenModel<T> AudioGen model for generating audio from text descriptions using neural audio codecs.
AudioGenResult<T> Result of audio generation.
AudioLDMClassifier<T> AudioLDM Classifier that repurposes AudioLDM’s latent representations for audio event detection.
AudioLM<T> AudioLM: audio generation as language modeling over semantic and acoustic tokens (Borsos et al.
AudioMAE<T> Audio-MAE (Masked Autoencoders for Audio) model for audio classification.
AudioSep<T> AudioSep - foundation model for open-vocabulary audio separation and sound event detection (Liu et al., ICML 2024).
AudioSuperResolution<T> Audio Super-Resolution model for upsampling low-resolution audio to high-resolution (Kuleshov et al., 2017; Li et al., 2021).
BEATs<T> BEATs (Audio Pre-Training with Acoustic Tokenizers) model for state-of-the-art audio event detection and classification.
BSRoFormer<T> BS-RoFormer (Band-Split Rotary Transformer) for music source separation.
BandSplitRNNEnhancer<T> Band-Split RNN speech enhancement model (Luo and Yu, 2023).
BandSplitRNN<T> BandSplitRNN for music source separation (Luo and Yu, 2023).
BasicPitch<T> Basic Pitch polyphonic music transcription model from Spotify.
BeatTracker<T> Extracts beat and tempo information from audio.
CAMPlusPlus<T> CAM++ (Context-Aware Masking Plus Plus) speaker verification model (Wang et al., 2023).
CLAPModel<T> CLAP (Contrastive Language-Audio Pretraining) — a dual-encoder neural network that learns to align audio and text representations in a shared embedding space via a contrastive objective.
CLAP<T> CLAP (Contrastive Language-Audio Pre-training) as an audio classifier: zero-shot against text prompts, or fine-tuned with a linear head.
CMGAN<T> CMGAN (Conformer-based Metric GAN) for speech enhancement.
CREPE<T> CREPE (Convolutional Representation for Pitch Estimation) neural pitch detector.
CRNNEventDetector<T> CRNN (Convolutional Recurrent Neural Network) model for Sound Event Detection.
CTCDecoder<T> CTC (Connectionist Temporal Classification) decoder-based speech recognition model.
Canary<T> Canary multilingual speech recognition and translation model from NVIDIA.
ChordRecognizer<T> Recognizes chords from audio using chromagram analysis.
ChordSegment<T> Represents a chord segment in audio.
ChromaExtractor<T> Extracts chromagram (pitch class profile) features from audio signals.
ChromaprintFingerprinter<T> Chromaprint-style audio fingerprinter based on chroma features.
Compressor<T> Dynamic range compressor effect.
ConformerFP<T> Conformer-based audio fingerprinting model combining self-attention with convolutions.
Conformer<T> Conformer (Gulati et al.
ConstantQTransform<T> Constant-Q Transform (CQT) for music analysis with logarithmic frequency resolution.
ConvTasNet<T> Conv-TasNet: A fully-convolutional time-domain audio separation network.
DAC<T> DAC, the Descript Audio Codec: a Snake-activated convolutional encoder–decoder with a residual vector quantizer of factorized, L2-normalized codes, trained with quantizer dropout against multi-period and multi-band multi-scale STFT discrimi…
DCCRN<T> DCCRN - Deep Complex Convolution Recurrent Network for speech enhancement.
DannaSep<T> Danna-Sep music source separation model using dual-path attention networks.
Data2Vec2<T> data2vec 2.0 self-supervised audio representation model.
DeepFilterNet<T> DeepFilterNet - State-of-the-art deep filtering network for speech enhancement.
DemucsNoise<T> Demucs for Noise - real-time noise suppression using the Demucs architecture (Defossez et al., 2020, Meta).
DiarizationResult Result of speaker diarization.
EAT<T> EAT (Efficient Audio Transformer) model for efficient audio event detection and classification.
ECAPATDNNLanguageIdentifier<T> ECAPA-TDNN (Emphasized Channel Attention, Propagation and Aggregation Time Delay Neural Network) for spoken language identification.
ECAPATDNNSpeaker<T> ECAPA-TDNN speaker verification and embedding extraction model.
Emotion2Vec<T> emotion2vec universal speech emotion recognition model.
EnCodec<T> EnCodec: a streaming convolutional encoder–decoder with an LSTM, residual vector quantization of its latent, and adversarial training against a multi-scale STFT discriminator with a loss balancer.
EncodedFrame<T> One encoded chunk: its codes [n_q, frames] and its volume scale (1 when not normalized).
EnergyBasedVad<T> Simple energy-based voice activity detector (algorithmic, no neural network).
EqBand<T> Represents a single EQ band with biquad filter.
FDYSED<T> FDY-SED (Frequency Dynamic Sound Event Detection) model for DCASE-winning SED.
FRCRN<T> FRCRN (Frequency Recurrence CRN) speech enhancement model (Zhao et al., ICASSP 2022).
FastConformer<T> Fast Conformer speech recognition model (Rekesh et al., 2023, NVIDIA NeMo).
FullSubNetPlus<T> FullSubNet+ (Full-Band and Sub-Band Fusion Network Plus) for speech enhancement.
GenreClassificationResult Result of genre classification.
GenreClassifier<T> Music genre classification after Tzanetakis and Cook 2002: a 30-dimensional “musical surface” feature vector per recording, classified by one diagonal-covariance Gaussian mixture per genre.
GenreFeatures Features extracted for genre classification.
HTDemucs<T> HTDemucs (Hybrid Transformer Demucs) for music source separation.
HTSAT<T> HTS-AT (Hierarchical Token-Semantic Audio Transformer) model for efficient audio classification.
HuBERTSER<T> HuBERT-SER (HuBERT for Speech Emotion Recognition) model.
HuBERT<T> HuBERT (Hidden-Unit BERT) self-supervised speech representation model.
IdentificationResult Result of speaker identification.
KeyDetectionResult<T> Result of key detection.
KeyDetector<T> Detects the musical key of audio using chromagram analysis.
LocalizationResult Result of sound source localization.
MERT<T> MERT self-supervised music understanding foundation model.
MPSENet<T> MP-SENet (Multi-Path Speech Enhancement Network) model (Lu et al., 2023).
MT3<T> MT3 (Multi-Track Music Transcription) model using T5-style encoder-decoder architecture.
MadmomBeatTracker<T> Madmom-style neural beat tracker using bidirectional RNNs.
MarbleNet<T> MarbleNet lightweight 1D separable convolutional VAD model (NVIDIA NeMo).
MelBandRoFormer<T> MelBand-RoFormer for mel-band music source separation.
MelodyExtractor<T> Neural Melody Extractor that identifies the primary melodic line from polyphonic audio.
MfccExtractor<T> Extracts Mel-Frequency Cepstral Coefficients (MFCCs) from audio signals.
MusicFlamingo<T> Music Flamingo multimodal music-language model for music understanding and reasoning.
MusicGenModel<T> MusicGen (Copet et al.
MusicGenTrainingSample<T> One MusicGen training example: a music clip, its text description and optional metadata.
MusicSegment<T> Represents a labeled segment of music structure.
MusicSourceSeparator<T> Music source separation model for separating audio into stems (vocals, drums, bass, other).
MusicStructureAnalyzer<T> Music Structure Analyzer that segments songs into structural sections (intro, verse, chorus, etc.).
MusicTaggingTransformer<T> Music Tagging Transformer for multi-label music tag prediction (genre, mood, instrument, era).
NeuralNoiseReducer<T> Neural network-based noise reducer for high-quality audio enhancement.
NeuralParametricEQ<T> Neural Parametric EQ model for automatic equalization (Steinmetz et al., 2022).
OnsetsAndFrames<T> Onsets and Frames piano transcription model from Google Magenta.
PANNsModel<T> PANNs (Pretrained Audio Neural Networks) audio classifier — a CNN14-style convolutional backbone over log-mel spectrograms, trained for AudioSet tagging (Kong et al.
PANNs<T> PANNs (Pre-trained Audio Neural Networks) CNN14 model for audio classification.
ParametricEqualizer<T> Multi-band parametric equalizer effect.
PeakNetFP<T> PeakNetFP spectral peak-based neural audio fingerprinting model.
Pengi<T> Pengi audio language model that frames all audio tasks as text-generation.
PyAnnote<T> pyannote 3.x end-to-end speaker diarization model.
QuailVad<T> Quail VAD - lightweight voice activity detection optimized for on-device deployment.
Qwen2Audio<T> Qwen2-Audio multimodal audio-language model for audio understanding and reasoning.
RNNTransducer<T> RNN-Transducer (RNN-T) streaming speech recognition model.
Result The decoded audio: interleaved samples in [−1, 1], the sample rate and the channel count.
Reverb<T> Algorithmic reverb effect using Schroeder-Moorer structure.
RoomImpulseResponse<T> FiNS — Filtered Noise Shaping network that estimates a time-domain room impulse response directly from reverberant speech.
SALMONN<T> SALMONN dual-encoder audio-language model for speech and audio understanding.
SCNet<T> SCNet (Sparse Compression Network) for music source separation (Tong et al., 2024).
SceneClassificationResult Result of acoustic scene classification.
SceneClassifier<T> Acoustic scene classification model for identifying recording environments.
SceneFeatures Features extracted for scene classification.
SeparationResult<T> Result of music source separation containing individual stems.
SileroVad<T> Silero Voice Activity Detection model - high accuracy neural network VAD.
SoundLocalizer<T> Sound source localization using microphone arrays.
SoundStream<T> SoundStream: a fully convolutional, causal encoder–decoder whose embeddings are compressed by a residual vector quantizer trained with quantizer dropout, so one model serves every bitrate up to its maximum, trained adversarially against a w…
SpeakerDiarizer<T> Performs speaker diarization (who spoke when) on audio recordings.
SpeakerEmbeddingExtractor<T> Extracts speaker embeddings (d-vectors) from audio for speaker recognition.
SpeakerEmbedding<T> Represents a speaker embedding vector.
SpeakerLM<T> SpeakerLM language-model-based speaker diarization and verification model.
SpeakerMatch A speaker match with score.
SpeakerTurn Represents a speaker turn in diarization output.
SpeakerVerifier<T> Verifies speaker identity by comparing embeddings against enrolled speakers.
SpectralFeatureExtractor<T> Extracts spectral features from audio signals including centroid, bandwidth, rolloff, and flux.
SpectralSubtractionEnhancer<T> Audio enhancer using spectral subtraction for noise reduction.
SpectrogramFingerprinter<T> Spectrogram peak-based audio fingerprinter (Shazam-style).
SpeechEmotionRecognizer<T> Neural network-based speech emotion recognition model that classifies emotional states from audio.
SpeechTokenizer<T> SpeechTokenizer: EnCodec’s encoder–quantizer–decoder with a BiLSTM encoder, whose first residual codebook is distilled toward a self-supervised teacher’s representations so that it carries the content of speech and the remaining codebooks t…
SpikingFullSubNet<T> Spiking-FullSubNet speech enhancement model using spiking neural networks.
StableAudioModel<T> Stable Audio model for generating high-quality audio from text descriptions.
TFGridNet<T> TF-GridNet (Time-Frequency GridNet) for speech enhancement and separation.
Tempogram<T> Neural Tempogram model for tempo estimation over time.
TitaNet<T> TitaNet speaker verification and embedding extraction model.
VerificationResult Result of speaker verification.
VoxLingua107Identifier<T> VoxLingua107 language identifier supporting 107 languages.
W2vBert<T> w2v-BERT: self-supervised speech representations from contrastive learning and masked prediction together (Chung et al.
Wav2Small<T> Wav2Small lightweight speech emotion recognition model (Gomez-Alanis et al., 2024).
Wav2Vec2LanguageIdentifier<T> Wav2Vec2 model fine-tuned for spoken language identification.
Wav2Vec2Model<T> Wav2Vec2 self-supervised speech recognition model.
Wav2Vec2<T> wav2vec 2.0 self-supervised speech representation model.
WavLMSpeaker<T> WavLM-based speaker verification and embedding extraction model (Chen et al., 2022).
WavLM<T> WavLM self-supervised speech representation model from Microsoft.
WebRTCVad<T> Neural WebRTC VAD model for low-latency voice activity detection.
WhisperModel<T> Whisper automatic speech recognition model for transcribing audio to text.
WhisperResult Result of Whisper transcription.
WhisperSegment A segment of transcribed speech with timing.
WhisperTokenizer Tokenizer for Whisper speech recognition model.
WhisperWord A transcribed word with timing information.
WorldPitchDetector<T> F0 estimation with WORLD’s DIO (Distributed Inline-filter Operation), refined by StoneMask.
YinPitchDetector<T> YIN pitch detection algorithm implementation.
YuE<T> YuE full-song music generation model with vocals and accompaniment.
Zipformer<T> Zipformer speech recognition model (Yao et al., 2023, Next-gen Kaldi).

Layers (2)

Type Summary
FactorizedVectorQuantizerLayer<T> DAC’s residual vector quantizer with factorized, L2-normalized codes (Kumar et al.
ResidualVectorQuantizerLayer<T> Residual vector quantization (SoundStream, Zeghidour et al.

Base Classes (11)

Type Summary
AudioClassifierBase<T> Base class for audio classification models (genre, event detection, scene classification).
AudioEffectBase<T> Base class for audio effects processors.
AudioEnhancerBase<T> Base class for algorithmic audio enhancement (non-neural network based).
AudioFeatureExtractorBase<T> Base class for audio feature extractors providing common functionality.
AudioFingerprinterBase<T> Base class for audio fingerprinting algorithms.
AudioNeuralNetworkBase<T> Base class for audio-focused neural networks that can operate in both ONNX inference and native training modes.
MusicAnalysisBase<T> Base class for music analysis algorithms (beat tracking, chord recognition, key detection).
NeuralAudioCodecBase<T> The shared pipeline of the neural audio codecs (SoundStream, EnCodec, DAC, SpeechTokenizer): an encoder to a latent sequence, a residual vector quantizer whose number of codebooks sets the bandwidth, and a decoder back to audio, trained adv…
PitchDetectorBase<T> Base class for pitch detection implementations.
SpeakerRecognitionBase<T> Base class for speaker recognition models (embedding extraction, verification, diarization).
VoiceActivityDetectorBase<T> Base class for algorithmic voice activity detection implementations (non-neural network).

Interfaces (1)

Type Summary
IResidualVectorQuantizer<T> A residual vector quantizer as the neural audio codecs use it: [1, dim, T] in, the sum of the code vectors of the active codebooks out, with codes per codebook.

Enums (13)

Type Summary
AudioGenModelSize Available AudioGen model sizes.
AudioLMStage Which of AudioLM’s three Transformers a training step updates.
EqFilterType Types of EQ filters.
KeyMode Musical key mode.
LocalizationAlgorithm Sound source localization algorithms.
MusicGenModelSize MusicGen’s model sizes (Copet et al.
MusicGenStereoPattern MusicGen’s stereo codebook patterns (Copet et al.
PANNsArchitecture PANNs architecture variants (legacy enum — the new code uses PANNsModelOptions with explicit dim / depth fields).
SoundStreamFilmPosition Where SoundStream applies its denoising FiLM conditioning (§III-F).
SpectralFeatureType Types of spectral features that can be extracted.
StableAudioModelSize Specifies the size variant of the Stable Audio model.
WhisperModelSize Available Whisper model sizes.
WindowType Window types for spectral analysis.

Options & Configuration (104)

Type Summary
ACEStepOptions Configuration options for the ACE-Step model.
ASTOptions Configuration options for the AST (Audio Spectrogram Transformer) models, AST and ASTModel, which run one network.
AudioEventDetectorOptions Options for audio event detection with Hershey et al.
AudioFlamingo2Options Configuration options for the Audio Flamingo 2 model.
AudioGenOptions Configuration options for audio generation models.
AudioLDMClassifierOptions Configuration options for the AudioLDM Classifier model.
AudioLMOptions Options for AudioLM (Borsos et al.
AudioMAEOptions Configuration options for the Audio-MAE (Masked Autoencoders for Audio) model.
AudioSepOptions Configuration options for the AudioSep (Audio Separation with Natural Language Queries) model.
AudioSuperResolutionOptions Configuration options for the Audio Super-Resolution model.
BEATsOptions Configuration options for the BEATs (Audio Pre-Training with Acoustic Tokenizers) model.
BSRoFormerOptions Configuration options for the BS-RoFormer (Band-Split Rotary Transformer) model.
BandSplitRNNEnhancerOptions Configuration options for the Band-Split RNN enhancement model.
BandSplitRNNOptions Configuration options for the BandSplitRNN source separation model.
BasicPitchOptions Configuration options for the Basic Pitch multi-pitch detection model.
BeatTrackerOptions Configuration options for beat tracking.
CAMPlusPlusOptions Configuration options for the CAM++ (Context-Aware Masking Plus Plus) speaker model.
CLAPOptions Configuration options for the CLAP (Contrastive Language-Audio Pre-training) model.
CMGANOptions Configuration options for the CMGAN (Conformer-based Metric GAN) speech enhancement model.
CREPEOptions Configuration options for the CREPE (Convolutional Representation for Pitch Estimation) model.
CRNNEventDetectorOptions Configuration options for the CRNN (Convolutional Recurrent Neural Network) Sound Event Detector.
CTCDecoderOptions Configuration options for the CTC Decoder speech recognition model.
CanaryOptions Configuration options for the Canary model.
ChordRecognizerOptions Configuration options for chord recognition.
ChromaOptions Options for chroma feature extraction.
ChromaprintOptions Configuration options for Chromaprint fingerprinting.
ConformerFPOptions Configuration options for the Conformer-based audio fingerprinting model.
ConformerOptions Options for Conformer.
DACOptions Options for DAC, the Descript Audio Codec (“High-Fidelity Audio Compression with Improved RVQGAN”, Kumar et al.
DannaSepOptions Configuration options for the Danna-Sep (Dual-path Attention Neural Network Audio Separator) model.
Data2Vec2Options Configuration options for the data2vec 2.0 self-supervised audio foundation model.
DemucsNoiseOptions Configuration options for the Demucs for Noise model.
EATOptions Configuration options for the EAT (Efficient Audio Transformer) model.
ECAPATDNNOptions Configuration options specific to ECAPA-TDNN language identification.
ECAPATDNNSpeakerOptions Configuration options for the ECAPA-TDNN speaker verification and embedding model.
Emotion2VecOptions Configuration options for the emotion2vec speech emotion recognition model.
EnCodecOptions Options for EnCodec (“High Fidelity Neural Audio Compression”, Défossez et al.
FDYSEDOptions Configuration options for the FDY-SED (Frequency Dynamic Sound Event Detection) model.
FRCRNOptions Configuration options for the FRCRN (Frequency Recurrence CRN) model.
FastConformerOptions Configuration options for the Fast Conformer speech recognition model.
FullSubNetPlusOptions Configuration options for the FullSubNet+ (Full-Band and Sub-Band Fusion Network Plus) model.
GenreClassifierOptions Options for genre classification after Tzanetakis and Cook 2002: 23 ms analysis windows, 1 s texture windows, 65536-sample beat histogram windows, and a three-component Gaussian mixture per genre.
GraFPrintOptions Configuration options for the GraFPrint graph-based audio fingerprinting model.
HTDemucsOptions Configuration options for the HTDemucs (Hybrid Transformer Demucs) model.
HTSATOptions Configuration options for the HTS-AT (Hierarchical Token-Semantic Audio Transformer) model.
HuBERTOptions Configuration options for the HuBERT (Hidden-Unit BERT) self-supervised speech model.
HuBERTSEROptions Configuration options for the HuBERT-based Speech Emotion Recognition model.
KeyDetectorOptions Configuration options for key detection.
LanguageIdentifierOptions Configuration options for language identification models.
MERTOptions Configuration options for the MERT music understanding foundation model.
MPSENetOptions Configuration options for the MP-SENet (Multi-Path Speech Enhancement Network) model.
MT3Options Configuration options for the MT3 (Multi-Track Music Transcription) model.
MadmomBeatTrackerOptions Configuration options for the Madmom-style neural beat tracker.
MarbleNetOptions Configuration options for the MarbleNet voice activity detection model.
MelBandRoFormerOptions Configuration options for the MelBand-RoFormer model.
MelodyExtractorOptions Configuration options for the neural Melody Extraction model.
MfccOptions Options for MFCC extraction.
MusicFlamingoOptions Configuration options for the Music Flamingo model.
MusicGenOptions Options for MusicGen (Copet et al.
MusicStructureAnalyzerOptions Configuration options for the Music Structure Analyzer model.
MusicTaggingTransformerOptions Configuration options for the Music Tagging Transformer model.
NeuralAudioCodecOptions Options shared by the neural audio codecs (encoder → residual vector quantizer → decoder).
NeuralFPOptions Configuration options for the Neural Audio Fingerprint (NeuralFP) model.
NeuralParametricEQOptions Configuration options for the Neural Parametric EQ model.
OnsetsAndFramesOptions Configuration options for the Onsets and Frames piano transcription model.
PANNsOptions Configuration options for the PANNs (Pre-trained Audio Neural Networks) model.
PeakNetFPOptions Configuration options for the PeakNetFP spectral peak-based fingerprinting model.
PengiOptions Configuration options for the Pengi model.
PyAnnoteOptions Configuration options for the pyannote 3.x speaker diarization model.
QuailVadOptions Configuration options for the Quail VAD model.
Qwen2AudioOptions Configuration options for the Qwen2-Audio multimodal audio-language model.
RNNTransducerOptions Configuration options for the RNN-Transducer (RNN-T) speech recognition model.
RoomImpulseResponseOptions Configuration options for the FiNS (Filtered Noise Shaping) room impulse response estimator.
SALMONNOptions Configuration options for the SALMONN multimodal audio-language model.
SCNetOptions Configuration options for the SCNet (Sparse Compression Network) source separation model.
SceneClassifierOptions Options for acoustic scene classification with the DCASE 2017 baseline system (Mesaros et al.
SoundLocalizerOptions Options for sound source localization.
SoundStreamOptions Options for SoundStream (“SoundStream: An End-to-End Neural Audio Codec”, Zeghidour et al.
SourceSeparationOptions Options for music source separation.
SpeakerDiarizerOptions Configuration options for speaker diarization.
SpeakerEmbeddingOptions Configuration options for the SpeakerEmbeddingExtractor.
SpeakerLMOptions Configuration options for the SpeakerLM model.
SpeakerVerifierOptions Configuration options for the SpeakerVerifier.
SpectralFeatureOptions Options for spectral feature extraction.
SpectrogramFingerprintOptions Configuration options for spectrogram fingerprinting.
SpeechEmotionRecognizerOptions Configuration options for the SpeechEmotionRecognizer.
SpeechTokenizerOptions Options for SpeechTokenizer (“SpeechTokenizer: Unified Speech Tokenizer for Speech Large Language Models”, Zhang et al., ICLR 2024).
SpikingFullSubNetOptions Configuration options for the Spiking-FullSubNet model.
StableAudioOptions Configuration options for Stable Audio generation.
TFGridNetOptions Configuration options for the TF-GridNet (Time-Frequency GridNet) speech enhancement model.
TempogramOptions Configuration options for the neural Tempogram tempo estimation model.
TitaNetOptions Configuration options for the TitaNet speaker verification and embedding model.
VoxLingua107Options Configuration options specific to VoxLingua107 language identification.
W2vBertOptions Options for w2v-BERT (Chung et al.
Wav2SmallOptions Configuration options for the Wav2Small speech emotion recognition model.
Wav2Vec2LidOptions Configuration options specific to Wav2Vec2 language identification.
Wav2Vec2Options Configuration options for the wav2vec 2.0 self-supervised speech model.
WavLMOptions Configuration options for the WavLM self-supervised speech model.
WavLMSEROptions Configuration options for the WavLM-SER speech emotion recognition model.
WavLMSpeakerOptions Configuration options for the WavLM Speaker verification and embedding model.
WebRTCVadOptions Configuration options for the WebRTC VAD neural model.
WhisperOptions Configuration options for the Whisper speech recognition model.
YuEOptions Configuration options for the YuE music generation model.
ZipformerOptions Configuration options for the Zipformer speech recognition model.

Helpers & Utilities (2)

Type Summary
Mp3Decoder An MPEG-1/2/2.5 Layer III (MP3) decoder: frame synchronization, side information, the bit reservoir, scale factors, Huffman decoding and requantization, mid/side and intensity stereo, reordering, alias reduction, the hybrid IMDCT and the po…
PitchWaveletTransform The pitch spectrogram of FastSpeech 2: a 10-scale continuous wavelet transform of the normalized log-F0 contour, and its approximate inverse.