Diffusion models have demonstrated remarkable success in generative tasks, including audio super-resolution (SR). In many applications like movie post-production and album mastering, substantial computational budgets are available for achieving superior audio quality. However, while existing diffusion approaches typically increase sampling steps to improve quality, the performance remains fundamentally limited by the stochastic nature of the sampling process, leading to high-variance and quality-limited outputs. Here, rather than simply increasing the number of sampling steps, we propose a different paradigm through inference-time scaling for SR, which explores multiple solution trajectories during the sampling process. Different task-specific verifiers are developed, and two search algorithms, including the random search and zero-order search for SR, are introduced. By actively guiding the exploration of the high-dimensional solution space through verifier-algorithm combinations, we enable more robust and higher-quality outputs. Through extensive validation across diverse audio domains (speech, music, sound effects) and frequency ranges, we demonstrate consistent performance gains, achieving improvements of up to 9.70% in aesthetics, 5.88% in speaker similarity, 15.20% in word error rate, and 46.98% in spectral distance for speech SR from 4 kHz to 24 kHz, showcasing the effectiveness of our approach.

Language-queried audio source separation (LASS) focuses on separating sounds using textual descriptions of the desired sources. Current methods mainly use discriminative approaches, such as time-frequency masking, to separate target sounds and minimize interference from other sources. However, these models face challenges when separating overlapping sound-tracks, which may lead to artifacts such as spectral holes or incomplete separation. Rectified flow matching (RFM), a generative model that establishes linear relations between the distribution of data and noise, offers superior theoretical properties and simplicity, but has not yet been explored in sound separation. In this work, we introduce FlowSep, a new generative model based on RFM for LASS tasks. FlowSep learns linear flow trajectories from noise to target source features within the variational autoencoder (VAE) latent space. During inference, the RFM-generated latent features are reconstructed into a mel-spectrogram via the pre-trained VAE decoder, followed by a pre-trained vocoder to synthesize the waveform. Trained on 1, 680 hours of audio data, FlowSep outperforms the state-of-the-art models across multiple benchmarks, as evaluated with subjective and objective metrics. Additionally, our results show that FlowSep surpasses a diffusion-based LASS model in both separation quality and inference efficiency, highlighting its strong potential for audio source separation tasks. Code, pre-trained models and demos can be found at: https://audio-agi.github.io/FlowSep_demo/.

While recent video-to-audio (V2A) models can generate realistic background audio from visual input, they largely overlook speech, an essential part of many video soundtracks. This paper proposes a new task, video-to-soundtrack (V2ST) generation, which aims to jointly produce synchronized background audio and speech within a unified framework. To tackle V2ST, we introduce DualDub, a unified framework built on a multimodal language model that integrates a multimodal encoder, a cross-modal aligner, and dual decoding heads for simultaneous background audio and speech generation. Specifically, our proposed cross-modal aligner employs causal and non-causal attention mechanisms to improve synchronization and acoustic harmony. Besides, to handle data scarcity, we design a curriculum learning strategy that progressively builds the multimodal capability. Finally, we introduce DualBench, the first benchmark for V2ST evaluation with a carefully curated test set and comprehensive metrics. Experimental results demonstrate that DualDub achieves state-of-the-art performance, generating high-quality and well-synchronized soundtracks with both speech and background audio. DualBench and generated samples of DualDub are available at https://github.com/wjtian-wonderful/DualBench.

Fish feeding intensity assessment (FFIA) aims to evaluate fish appetite changes during feeding, which is crucial in industrial aquaculture applications. Existing FFIA methods are limited by their robustness to noise, computational complexity, and the lack of public datasets for developing the models. To address these issues, we first introduce AV-FFIA, a new dataset containing 27,000 labeled audio and video clips that capture different levels of fish feeding intensity. Then, we introduce multi-modal approaches for FFIA by leveraging the models pre-trained on individual modalities and fused with data fusion methods. We perform benchmark studies of these methods on AV-FFIA, and demonstrate the advantages of the multi-modal approach over the single-modality based approach, especially in noisy environments. However, compared to the methods developed for individual modalities, the multimodal approaches may involve higher computational costs due to the need for independent encoders for each modality. To overcome this issue, we further present a novel unified mixed-modality based method for FFIA, termed as U-FFIA. U-FFIA is a single model capable of processing audio, visual, or audio-visual modalities, by leveraging modality dropout during training and knowledge distillation using the models pre-trained with data from single modality. We demonstrate that U-FFIA can achieve performance better than or on par with the state-of-the-art modality-specific FFIA models, with significantly lower computational overhead, enabling robust and efficient FFIA for improved aquaculture management. To encourage further research, we have released the AV-FFIA dataset, the pre-trained model and codes athttps://github.com/FishMaster93/U-FFIA. Note to Practitioners—Feeding is one of the most important costs in aquaculture. However, current feeding machines usually operate with fixed thresholds or human experiences, lacking the ability to automatically adjust to fish feeding intensity. FFIA can evaluate the intensity changes in fish appetite during the feeding process and optimize the control strategies of the feeding machine to avoid inadequate feeding or overfeeding, thereby reducing the feeding cost and improving the well-being of fish in industrial aquaculture. The existing methods have mainly exploited single-modality data, and have a high sensitivity to input noise. Using video and audio offers improved chances to address the challenges brought by various environments. However, compared with processing data from single modalities, using multiple modalities simultaneously often involves increased computational resources, including memory, processing power, and storage. This can impact system performance and scalability. To address these issues, we focus on the efficient unified model, which is capable of processing both multimodal and single-modal input. Our proposed model achieved state-of-the-art (SOTA) performance in FFIA with high computational efficiency.

The audio spectrogram is a time-frequency representation that has been widely used for audio classification. One of the key attributes of the audio spectrogram is the temporal resolution, which depends on the hop size used in the Short-Time Fourier Transform (STFT). Previous works generally assume the hop size should be a constant value (e.g., 10 ms). However, a fixed temporal resolution is not always optimal for different types of sound. The temporal resolution affects not only classification accuracy but also computational cost. This paper proposes a novel method, DiffRes, that enables differentiable temporal resolution modeling for audio classification. Given a spectrogram calculated with a fixed hop size, DiffRes merges non-essential time frames while preserving important frames. DiffRes acts as a "drop-in" module between an audio spectrogram and a classifier and can be jointly optimized with the classification task. We evaluate DiffRes on five audio classification tasks, using mel-spectrograms as the acoustic features, followed by off-the-shelf classifier backbones. Compared with previous methods using the fixed temporal resolution, the DiffRes-based method can achieve the equivalent or better classification accuracy with at least 25% computational cost reduction. We further show that DiffRes can improve classification accuracy by increasing the temporal resolution of input acoustic features, without adding to the computational cost.

Diffusion models have shown promising results in cross-modal generation tasks, including text-to-image and text-to-audio generation. However, generating music, as a special type of audio, presents unique challenges due to limited availability of music data and sensitive issues related to copyright and plagiarism. In this paper, to tackle these challenges, we first construct a state-of-the-art text-to-music model, MusicLDM, that adapts Stable Diffusion and AudioLDM architectures to the music domain. Then, to address the limitations of training data and to avoid plagiarism, we leverage a beat tracking model and propose two different mixup strategies for data augmentation: beat-synchronous audio mixup and beat-synchronous latent mixup, which recombine training audio directly or via a latent embeddings space, respectively. Such mixup strategies encourage the model to interpolate between musical training samples and generate new music within the convex hull of the training data, making the generated music more diverse while still staying faithful to the corresponding style. In addition to popular evaluation metrics, we design several new evaluation metrics based on CLAP score to demonstrate that our proposed MusicLDM and beat-synchronous mixup strategies improve both the quality and novelty of generated music, as well as the correspondence between input text and generated music.

Audio super-resolution is a fundamental task that predicts high-frequency components for low-resolution audio, enhancing audio quality in digital applications. Previous methods have limitations such as the limited scope of audio types (e.g., music, speech) and specific bandwidth settings they can handle (e.g., 4 kHz to 8 kHz). In this paper, we introduce a diffusion-based generative model, AudioSR, that is capable of performing robust audio super-resolution on versatile audio types, including sound effects, music, and speech. Specifically, AudioSR can upsample any input audio signal within the bandwidth range of 2 kHz to 16 kHz to a high-resolution audio signal at 24 kHz bandwidth with a sampling rate of 48 kHz. Extensive objective evaluation on various audio super-resolution benchmarks demonstrates the strong result achieved by the proposed model. In addition, our subjective evaluation shows that AudioSR can act as a plug-and-play module to enhance the generation quality of a wide range of audio generative models, including AudioLDM, Fastspeech2, and MusicGen. Our code and demo are available at https://audioldm.github.io/audiosr.

Despite recent progress in text-to-audio (TTA) generation, we show that the state-of-the-art models, such as AudioLDM, trained on datasets with an imbalanced class distribution, such as AudioCaps, are biased in their generation performance. Specifically, they excel in generating common audio classes while underperforming in the rare ones, thus degrading the overall generation performance. We refer to this problem as long-tailed text-to-audio generation. To address this issue, we propose a simple retrieval-augmented approach for TTA models. Specifically, given an input text prompt, we first leverage a Contrastive Language Audio Pretraining (CLAP) model to retrieve relevant text-audio pairs. The features of the retrieved audio-text data are then used as additional conditions to guide the learning of TTA models. We enhance AudioLDM with our proposed approach and denote the resulting augmented system as Re-AudioLDM. On the AudioCaps dataset, Re-AudioLDM achieves a state-of-the-art Frechet Audio Distance (FAD) of 1.37, outperforming the existing approaches by a large margin. Furthermore, we show that Re-AudioLDM can generate realistic audio for complex scenes, rare audio classes, and even unseen audio types, indicating its potential in TTA tasks.

First-shot (FS) unsupervised anomalous sound detection (ASD) is a brand-new task introduced in DCASE 2023 Challenge Task 2, where the anomalous sounds for the target machine types are unseen in training. Existing methods often rely on the availability of normal and abnormal sound data from the target machines. However, due to the lack of anomalous sound data for the target machine types, it becomes challenging when adapting the existing ASD methods to the first-shot task. In this paper, we propose a new framework for the first-shot unsupervised ASD, where metadata-assisted audio generation is used to estimate unknown anomalies, by utilising the available machine information (i.e., metadata and sound data) to fine-tune a text-to-audio generation model for generating the anomalous sounds that contain unique acoustic characteristics accounting for each different machine type. We then use the method of Time-Weighted Frequency domain audio Representation with Gaussian Mixture Model (TWFRGMM) as the backbone to achieve the first-shot unsupervised ASD. Our proposed FS-TWFR-GMM method achieves competitive performance amongst top systems in DCASE 2023 Challenge Task 2, while requiring only 1% model parameters for detection, as validated in our experiments.

Recent progress in large-scale zero-shot speech synthesis has been significantly advanced by language models and diffusion models. However, the generation process of both methods is slow and computationally intensive. Efficient speech synthesis using a lower computing budget to achieve quality on par with previous work remains a significant challenge. In this paper, we present FlashSpeech, a large-scale zero-shot speech synthesis system with approximately 5% of the inference time compared with previous work. FlashSpeech is built on the latent consistency model and applies a novel adversarial consistency training approach that can train from scratch without the need for a pre-trained diffusion model as the teacher. Furthermore, a new prosody generator module enhances the diversity of prosody, making the rhythm of the speech sound more natural. The generation processes of FlashSpeech can be achieved efficiently with one or two sampling steps while maintaining high audio quality and high similarity to the audio prompt for zero-shot speech generation. Our experimental results demonstrate the superior performance of FlashSpeech. Notably, FlashSpeech can be about 20 times faster than other zero-shot speech synthesis systems while maintaining comparable performance in terms of voice quality and similarity. Furthermore, FlashSpeech demonstrates its versatility by efficiently performing tasks like voice conversion, speech editing, and diverse speech sampling. Audio samples can be found in https://flashspeech.github.io/

Text-to-speech (TTS) has made rapid progress in both academia and industry in recent years. Some questions naturally arise that whether a TTS system can achieve human-level quality, how to define/judge that quality, and how to achieve it. In this paper, we answer these questions by first defining the human-level quality based on the statistical significance of subjective measure and introducing appropriate guidelines to judge it, and then developing a TTS system called NaturalSpeech that achieves human-level quality on benchmark datasets. Specifically, we leverage a variational auto-encoder (VAE) for end-to-end text-to-waveform generation, with several key modules to enhance the capacity of the prior from text and reduce the complexity of the posterior from speech, including phoneme pre-training, differentiable duration modeling, bidirectional prior/posterior modeling, and a memory mechanism in VAE. Experimental evaluations on the popular LJSpeech dataset show that our proposed NaturalSpeech achieves -0.01 CMOS (comparative mean opinion score) to human recordings at the sentence level, with Wilcoxon signed rank test at p-level p >> 0.05, which demonstrates no statistically significant difference from human recordings for the first time.

Although audio generation shares commonalities across different types of audio, such as speech, music, and sound effects, designing models for each type requires careful consideration of specific objectives and biases that can significantly differ from those of other types. To bring us closer to a unified perspective of audio generation, this paper proposes a holistic framework that utilizes the same learning method for speech, music, and sound effect generation. Our framework utilizes a general representation of audio, called “language of audio” (LOA). Any audio can be translated into LOA based on AudioMAE, a self-supervised pre-trained representation learning model. In the generation process, we translate other modalities into LOA by using a GPT-2 model, and we perform self-supervised audio generation learning with a latent diffusion model conditioned on the LOA of audio in our training set. The proposed framework naturally brings advantages such as reusable self-supervised pretrained latent diffusion models. Experiments on the major benchmarks of text-to-audio, text-to-music, and text-to-speech with three AudioLDM 2 variants demonstrate competitive performance of the AudioLDM 2 variants framework against previous approaches. Our code, pretrained model, and demo are available athttps://audioldm.github.io/audioldm2.

The advancement of audio-language (AL) multimodal learning tasks has been significant in recent years, yet the limited size of existing audio-language datasets poses challenges for researchers due to the costly and time-consuming collection process. To address this data scarcity issue, we introduceWavCaps, the first large-scale weakly-labelled audio captioning dataset, comprising approximately 400 k audio clips with paired captions. We sourced audio clips and their raw descriptions from web sources and a sound event detection dataset. However, the online-harvested raw descriptions are highly noisy and unsuitable for direct use in tasks such as automated audio captioning. To overcome this issue, we propose a three-stage processing pipeline for filtering noisy data and generating high-quality captions, where ChatGPT, a large language model, is leveraged to filter and transform raw descriptions automatically. We conduct a comprehensive analysis of the characteristics of WavCaps dataset and evaluate it on multiple downstream audio-language multimodal learning tasks. The systems trained on WavCaps outperform previous state-of-the-art (SOTA) models by a significant margin. Our aspiration is for the WavCaps dataset we have proposed to facilitate research in audio-language multimodal learning and demonstrate the potential of utilizing large language models (LLMs) to enhance academic research.

Recently, there has been increasing interest in building efficient audio neural networks for on-device scenarios. Most existing approaches are designed to reduce the size of audio neural networks using methods such as model pruning. In this work, we show that instead of reducing model size using complex methods, eliminating the temporal redundancy in the input audio features (e.g., mel-spectrogram) could be an effective approach for efficient audio classification. To do so, we proposed a family of simple pooling front-ends (SimPFs) which use simple non-parametric pooling operations to reduce the redundant information within the mel-spectrogram. We perform extensive experiments on four audio classification tasks to evaluate the performance of SimPFs. Experimental results show that SimPFs can achieve a reduction in more than half of the number of floating point operations (FLOPs) for off-the-shelf audio neural networks, with negligible degradation or even some improvements in audio classification performance.

Text-to-audio (TTA) systems have recently gained attention for their ability to synthesize general audio based on text descriptions. However, previous studies in TTA have limited generation quality with high computational costs. In this study, we propose AudioLDM, a TTA system that is built on a latent space to learn continuous audio representations from contrastive language-audio pretraining (CLAP) embeddings. The pretrained CLAP models enable us to train LDMs with audio embeddings while providing text embeddings as the condition during sampling. By learning the latent representations of audio signals without modelling the cross-modal relationship, AudioLDM improves both generation quality and computational efficiency. Trained on AudioCaps with a single GPU, AudioLDM achieves state-of-the-art TTA performance compared to other open-sourced systems, measured by both objective and subjective metrics. AudioLDM is also the first TTA system that enables various text-guided audio manipulations (e.g., style transfer) in a zero-shot fashion. Our implementation and demos are available at https://audioldm.github.io.

Contrastive language-audio pretraining (CLAP) has become a new paradigm to learn audio concepts with audio-text pairs. CLAP models have shown unprecedented performance as zero-shot classifiers on downstream tasks. To further adapt CLAP with domain-specific knowledge, a popular method is to finetune its audio encoder with available labelled examples. However, this is challenging in low-shot scenarios, as the amount of annotations is limited compared to the model size. In this work, we introduce a Training-efficient (Treff) adapter to rapidly learn with a small set of examples while maintaining the capacity for zero-shot classification. First, we propose a cross-attention linear model (CALM) to map a set of labelled examples and test audio to test labels. Second, we find initialising CALM as a cosine measurement improves our Treff adapter even without training. The Treff adapter outperforms metric-based methods in few-shot settings and yields competitive results to fully-supervised methods.

Audio captioning aims to generate text descriptions of audio clips.In the real world, many objects produce similar sounds.How to accurately recognize ambiguous sounds is a major challenge for audio captioning.In this work, inspired by inherent human multimodal perception, we propose visuallyaware audio captioning, which makes use of visual information to help the description of ambiguous sounding objects.Specifically, we introduce an off-the-shelf visual encoder to extract video features and incorporate the visual features into an audio captioning system.Furthermore, to better exploit complementary audio-visual contexts, we propose an audio-visual attention mechanism that adaptively integrates audio and visual context and removes the redundant information in the latent space.Experimental results on AudioCaps, the largest audio captioning dataset, show that our proposed method achieves state-of-theart results on machine translation metrics.

This study defines a new evaluation metric for audio tagging tasks to alleviate the limitation of the mean average precision (mAP) metric.The mAP metric treats different kinds of sound as independent classes without considering their relations.The proposed metric, ontology-aware mean average precision (OmAP), addresses the weaknesses of mAP by utilizing additional ontology during evaluation.Specifically, we reweight the false positive events in the model prediction based on the AudioSet ontology graph distance to the target classes.The OmAP also provides insights into model performance by evaluating different coarse-grained levels in the ontology graph.We conduct a human assessment and show that OmAP is more consistent with human perception than mAP.We also propose an ontology-based loss function (OBCE) that reweights binary cross entropy (BCE) loss based on the ontology distance.Our experiment shows that OBCE can improve both mAP and OmAP metrics on the AudioSet tagging task.

Speech super-resolution (SR) is a task to increase speech sampling rate by generating high-frequency components. Existing speech SR methods are trained in constrained experimental settings, such as a fixed upsampling ratio. These strong constraints can potentially lead to poor generalization ability in mismatched real-world cases. In this paper, we propose a neural vocoder based speech super-resolution method (NVSR) that can handle a variety of input resolution and upsampling ratios. NVSR consists of a mel-bandwidth extension module, a neural vocoder module, and a post-processing module. Our proposed system achieves state-of-the-art results on the VCTK multi-speaker benchmark. On 44.1 kHz target resolution, NVSR outperforms WSRGlow and Nu-wave by 8% and 37% respectively on log spectral distance and achieves a significantly better perceptual quality. We also demonstrate that prior knowledge in the pre-trained vocoder is crucial for speech SR by performing mel-bandwidth extension with a simple replication-padding method. Samples can be found in https://haoheliu.github.io/nvsr.

In this paper, we introduce the task of language-queried audio source separation (LASS), which aims to separate a target source from an audio mixture based on a natural language query of the target source (e.g., "a man tells a joke followed by people laughing"). A unique challenge in LASS is associated with the complexity of natural language description and its relation with the audio sources. To address this issue, we proposed LASS-Net, an end-to-end neural network that is learned to jointly process acoustic and linguistic information, and separate the target source that is consistent with the language query from an audio mixture. We evaluate the performance of our proposed system with a dataset created from the AudioCaps dataset. Experimental results show that LASS-Net achieves considerable improvements over baseline methods. Furthermore, we observe that LASS-Net achieves promising generalization results when using diverse human-annotated descriptions as queries, indicating its potential use in real-world scenarios. The separated audio samples and source code are available at https://liuxubo717.github.io/LASS-demopage.

Speech restoration aims to remove distortions in speech signals. Prior methods mainly focus on a single type of distortion, such as speech denoising or dereverberation. However, speech signals can be degraded by several different distortions simultaneously in the real world. It is thus important to extend speech restoration models to deal with multiple distortions. In this paper, we introduce VoiceFixer, a unified framework for high-fidelity speech restoration. VoiceFixer restores speech from multiple distortions (e.g., noise, reverberation, and clipping) and can expand degraded speech (e.g., noisy speech) with a low bandwidth to 44.1 kHz full-bandwidth high-fidelity speech. We design VoiceFixer based on (1) an analysis stage that predicts intermediate-level features from the degraded speech, and (2) a synthesis stage that generates waveform using a neural vocoder. Both objective and subjective evaluations show that VoiceFixer is effective on severely degraded speech, such as real-world historical speech recordings. Samples of VoiceFixer are available at https://haoheliu.github.io/voicefixer.

Binaural audio plays a significant role in constructing immersive augmented and virtual realities. As it is expensive to record binaural audio from the real world, synthesizing them from mono audio has attracted increasing attention. This synthesis process involves not only the basic physical warping of the mono audio, but also room reverberations and head/ear related filtration, which, however, are difficult to accurately simulate in traditional digital signal processing. In this paper, we formulate the synthesis process from a different perspective by decomposing the binaural audio into a common part that shared by the left and right channels as well as a specific part that differs in each channel. Accordingly, we propose BinauralGrad, a novel two-stage framework equipped with diffusion models to synthesize them respectively. Specifically, in the first stage, the common information of the binaural audio is generated with a single-channel diffusion model conditioned on the mono audio, based on which the binaural audio is generated by a two-channel diffusion model in the second stage. Combining this novel perspective of two-stage synthesis with advanced generative models (i.e., the diffusion models), the proposed BinauralGrad is able to generate accurate and high-fidelity binaural audio samples. Experiment results show that on a benchmark dataset, BinauralGrad outperforms the existing baselines by a large margin in terms of both object and subject evaluation metrics (Wave L2: $0.128$ vs. $0.157$, MOS: $3.80$ vs. $3.61$). The generated audio samples\footnote{\url{https://speechresearch.github.io/binauralgrad}} and code\footnote{\url{https://github.com/microsoft/NeuralSpeech/tree/master/BinauralGrad}} are available online.

Speech enhancement is a task to improve the intelligibility and perceptual quality of degraded speech signal.Recently, neural networks based methods have been applied to speech enhancement.However, many neural network based methods require noisy and clean speech pairs for training.We propose a speech enhancement framework that can be trained with large-scale weakly labelled AudioSet dataset.Weakly labelled data only contain audio tags of audio clips, but not the onset or offset times of speech.We first apply pretrained audio neural networks (PANNs) to detect anchor segments that contain speech or sound events in audio clips.Then, we randomly mix two detected anchor segments containing speech and sound events as a mixture, and build a conditional source separation network using PANNs predictions as soft conditions for speech enhancement.In inference, we input a noisy speech signal with the one-hot encoding of "Speech" as a condition to the trained system to predict enhanced speech.Our system achieves a PESQ of 2.28 and an SSNR of 8.75 dB on the VoiceBank-DEMAND dataset, outperforming the previous SEGAN system of 2.16 and 7.73 dB respectively.

This paper presents a new input format, channel-wise subband input (CWS), for convolutional neural networks (CNN) based music source separation (MSS) models in the frequency domain. We aim to address the major issues in CNN-based high-resolution MSS model: high computational cost and weight sharing between distinctly different bands. Specifically, in this paper, we decompose the input mixture spectra into several bands and concatenate them channel-wise as the model input. The proposed approach enables effective weight sharing in each subband and introduces more flexibility between channels. For comparison purposes, we perform voice and accompaniment separation (VAS) on models with different scales, architectures, and CWS settings. Experiments show that the CWS input is beneficial in many aspects. We evaluate our method on musdb18hq test set, focusing on SDR, SIR and SAR metrics. Among all our experiments, CWS enables models to obtain 6.9% performance gain on the average metrics. With even a smaller number of parameters, less training data, and shorter training time, our MDenseNet with 8-bands CWS input still surpasses the original MMDenseNet with a large margin. Moreover, CWS also reduces computational cost and training time to a large extent.