This paper proposes SEFGAN, a Deep Neural Network (DNN) combining maximum likelihood training and Generative Adversarial Networks (GANs) for efficient speech enhancement (SE). For this, a DNN is trained to synthesize the enhanced speech conditioned on noisy speech using a Normalizing Flow (NF) as generator in a GAN framework. While the combination of likelihood models and GANs is not trivial, SEFGAN demonstrates that a hybrid adversarial and maximum likelihood training approach enables the model to maintain high quality audio generation and log-likelihood estimation. Our experiments indicate that this approach strongly outperforms the baseline NF-based model without introducing additional complexity to the enhancement network. A comparison using computational metrics and a listening experiment reveals that SEFGAN is competitive with other state-of-the-art models.
Language models have been successfully used to model natural signals, such as images, speech, and music. A key component of these models is a high quality neural compression model that can compress high-dimensional natural signals into lower dimensional discrete tokens. To that end, we introduce a high-fidelity universal neural audio compression algorithm that achieves ~90x compression of 44.1 KHz audio into tokens at just 8kbps bandwidth. We achieve this by combining advances in high-fidelity audio generation with better vector quantization techniques from the image domain, along with improved adversarial and reconstruction losses. We compress all domains (speech, environment, music, etc.) with a single universal model, making it widely applicable to generative modeling of all audio. We compare with competing audio compression algorithms, and find our method outperforms them significantly. We provide thorough ablations for every design choice, as well as open-source code and trained model weights. We hope our work can lay the foundation for the next generation of high-fidelity audio modeling.
We present Composable Diffusion (CoDi), a novel generative model capable of generating any combination of output modalities, such as language, image, video, or audio, from any combination of input modalities. Unlike existing generative AI systems, CoDi can generate multiple modalities in parallel and its input is not limited to a subset of modalities like text or image. Despite the absence of training datasets for many combinations of modalities, we propose to align modalities in both the input and output space. This allows CoDi to freely condition on any input combination and generate any group of modalities, even if they are not present in the training data. CoDi employs a novel composable generation strategy which involves building a shared multimodal space by bridging alignment in the diffusion process, enabling the synchronized generation of intertwined modalities, such as temporally aligned video and audio. Highly customizable and flexible, CoDi achieves strong joint-modality generation quality, and outperforms or is on par with the unimodal state-of-the-art for single-modality synthesis. The project page with demonstrations and code is at https://codi-gen.github.io
Contrastive Language-Audio Pretraining (CLAP) became of crucial importance in the field of audio and speech processing. Its employment ranges from sound event detection to text-to-audio generation. However, one of the main limitations is the considerable amount of data required in the training process and the overall computational complexity during inference. This paper investigates how we can reduce the complexity of contrastive language-audio pre-trained models, yielding an efficient model that we call tinyCLAP. We derive an unimodal distillation loss from first principles and explore how the dimensionality of the shared, multimodal latent space can be reduced via pruning. TinyCLAP uses only 6% of the original Microsoft CLAP parameters with a minimal reduction (less than 5%) in zero-shot classification performance across the three sound event detection datasets on which it was tested
Binaural stereo audio is recorded by imitating the way the human ear receives sound, which provides people with an immersive listening experience. Existing approaches leverage autoencoders and directly exploit visual spatial information to synthesize binaural stereo, resulting in a limited representation of visual guidance. For the first time, we propose a visually guided generative adversarial approach for generating binaural stereo audio from mono audio. Specifically, we develop a Stereo Audio Generation Model (SAGM), which utilizes shared spatio-temporal visual information to guide the generator and the discriminator to work separately. The shared visual information is updated alternately in the generative adversarial stage, allowing the generator and discriminator to deliver their respective guided knowledge while visually sharing. The proposed method learns bidirectional complementary visual information, which facilitates the expression of visual guidance in generation. In addition, spatial perception is a crucial attribute of binaural stereo audio, and thus the evaluation of stereo spatial perception is essential. However, previous metrics failed to measure the spatial perception of audio. To this end, a metric to measure the spatial perception of audio is proposed for the first time. The proposed metric is capable of measuring the magnitude and direction of spatial perception in the temporal dimension. Further, considering its function, it is feasible to utilize it instead of demanding user studies to some extent. The proposed method achieves state-of-the-art performance on 2 datasets and 5 evaluation metrics. Qualitative experiments and user studies demonstrate that the method generates space-realistic stereo audio.
Text-to-audio generation (TTA) produces audio from a text description, learning from pairs of audio samples and hand-annotated text. However, commercializing audio generation is challenging as user-input prompts are often under-specified when compared to text descriptions used to train TTA models. In this work, we treat TTA models as a ``blackbox'' and address the user prompt challenge with two key insights: (1) User prompts are generally under-specified, leading to a large alignment gap between user prompts and training prompts. (2) There is a distribution of audio descriptions for which TTA models are better at generating higher quality audio, which we refer to as ``audionese''. To this end, we rewrite prompts with instruction-tuned models and propose utilizing text-audio alignment as feedback signals via margin ranking learning for audio improvements. On both objective and subjective human evaluations, we observed marked improvements in both text-audio alignment and music audio quality.
Distributional shift is a central challenge in the deployment of machine learning models as they can be ill-equipped for real-world data. This is particularly evident in text-to-audio generation where the encoded representations are easily undermined by unseen prompts, which leads to the degradation of generated audio -- the limited set of the text-audio pairs remains inadequate for conditional audio generation in the wild as user prompts are under-specified. In particular, we observe a consistent audio quality degradation in generated audio samples with user prompts, as opposed to training set prompts. To this end, we present a retrieval-based in-context prompt editing framework that leverages the training captions as demonstrative exemplars to revisit the user prompts. We show that the framework enhanced the audio quality across the set of collected user prompts, which were edited with reference to the training captions as exemplars.
In this paper, we explore audio-editing with non-rigid text edits. We show that the proposed editing pipeline is able to create audio edits that remain faithful to the input audio. We explore text prompts that perform addition, style transfer, and in-painting. We quantitatively and qualitatively show that the edits are able to obtain results which outperform Audio-LDM, a recently released text-prompted audio generation model. Qualitative inspection of the results points out that the edits given by our approach remain more faithful to the input audio in terms of keeping the original onsets and offsets of the audio events.
Designing effective automatic speech recognition (ASR) systems for Code-Switching (CS) often depends on the availability of the transcribed CS resources. To address data scarcity, this paper introduces Speech Collage, a method that synthesizes CS data from monolingual corpora by splicing audio segments. We further improve the smoothness quality of audio generation using an overlap-add approach. We investigate the impact of generated data on speech recognition in two scenarios: using in-domain CS text and a zero-shot approach with synthesized CS text. Empirical results highlight up to 34.4% and 16.2% relative reductions in Mixed-Error Rate and Word-Error Rate for in-domain and zero-shot scenarios, respectively. Lastly, we demonstrate that CS augmentation bolsters the model's code-switching inclination and reduces its monolingual bias.
Diffusion models have gained prominence in the image domain for their capabilities in data generation and transformation, achieving state-of-the-art performance in various tasks in both image and audio domains. In the rapidly evolving field of audio-based machine learning, safeguarding model integrity and establishing data copyright are of paramount importance. This paper presents the first watermarking technique applied to audio diffusion models trained on mel-spectrograms. This offers a novel approach to the aforementioned challenges. Our model excels not only in benign audio generation, but also incorporates an invisible watermarking trigger mechanism for model verification. This watermark trigger serves as a protective layer, enabling the identification of model ownership and ensuring its integrity. Through extensive experiments, we demonstrate that invisible watermark triggers can effectively protect against unauthorized modifications while maintaining high utility in benign audio generation tasks.
Recent advancements in audio generation have been spurred by the evolution of large-scale deep learning models and expansive datasets. However, the task of video-to-audio (V2A) generation continues to be a challenge, principally because of the intricate relationship between the high-dimensional visual and auditory data, and the challenges associated with temporal synchronization. In this study, we introduce FoleyGen, an open-domain V2A generation system built on a language modeling paradigm. FoleyGen leverages an off-the-shelf neural audio codec for bidirectional conversion between waveforms and discrete tokens. The generation of audio tokens is facilitated by a single Transformer model, which is conditioned on visual features extracted from a visual encoder. A prevalent problem in V2A generation is the misalignment of generated audio with the visible actions in the video. To address this, we explore three novel visual attention mechanisms. We further undertake an exhaustive evaluation of multiple visual encoders, each pretrained on either single-modal or multi-modal tasks. The experimental results on VGGSound dataset show that our proposed FoleyGen outperforms previous systems across all objective metrics and human evaluations.
Diffusion models are instrumental in text-to-audio (TTA) generation. Unfortunately, they suffer from slow inference due to an excessive number of queries to the underlying denoising network per generation. To address this bottleneck, we introduce ConsistencyTTA, a framework requiring only a single non-autoregressive network query, thereby accelerating TTA by hundreds of times. We achieve so by proposing "CFG-aware latent consistency model," which adapts consistency generation into a latent space and incorporates classifier-free guidance (CFG) into model training. Moreover, unlike diffusion models, ConsistencyTTA can be finetuned closed-loop with audio-space text-aware metrics, such as CLAP score, to further enhance the generations. Our objective and subjective evaluation on the AudioCaps dataset shows that compared to diffusion-based counterparts, ConsistencyTTA reduces inference computation by 400x while retaining generation quality and diversity.
This paper presents an innovative approach to enhance control over audio generation by emphasizing the alignment between audio and text representations during model training. In the context of language model-based audio generation, the model leverages input from both textual and audio token representations to predict subsequent audio tokens. However, the current configuration lacks explicit regularization to ensure the alignment between the chosen text representation and the language model's predictions. Our proposal involves the incorporation of audio and text representation regularization, particularly during the classifier-free guidance (CFG) phase, where the text condition is excluded from cross attention during language model training. The aim of this proposed representation regularization is to minimize discrepancies in audio and text similarity compared to other samples within the same training batch. Experimental results on both music and audio generation tasks demonstrate that our proposed methods lead to improvements in objective metrics for both audio and music generation, as well as an enhancement in the human perception for audio generation.
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.
Sampling from known probability distributions is a ubiquitous task in computational science, underlying calculations in domains from linguistics to biology and physics. Generative machine-learning (ML) models have emerged as a promising tool in this space, building on the success of this approach in applications such as image, text, and audio generation. Often, however, generative tasks in scientific domains have unique structures and features -- such as complex symmetries and the requirement of exactness guarantees -- that present both challenges and opportunities for ML. This Perspective outlines the advances in ML-based sampling motivated by lattice quantum field theory, in particular for the theory of quantum chromodynamics. Enabling calculations of the structure and interactions of matter from our most fundamental understanding of particle physics, lattice quantum chromodynamics is one of the main consumers of open-science supercomputing worldwide. The design of ML algorithms for this application faces profound challenges, including the necessity of scaling custom ML architectures to the largest supercomputers, but also promises immense benefits, and is spurring a wave of development in ML-based sampling more broadly. In lattice field theory, if this approach can realize its early promise it will be a transformative step towards first-principles physics calculations in particle, nuclear and condensed matter physics that are intractable with traditional approaches.
Recent breakthroughs in zero-shot voice synthesis have enabled imitating a speaker's voice using just a few seconds of recording while maintaining a high level of realism. Alongside its potential benefits, this powerful technology introduces notable risks, including voice fraud and speaker impersonation. Unlike the conventional approach of solely relying on passive methods for detecting synthetic data, watermarking presents a proactive and robust defence mechanism against these looming risks. This paper introduces an innovative audio watermarking framework that encodes up to 32 bits of watermark within a mere 1-second audio snippet. The watermark is imperceptible to human senses and exhibits strong resilience against various attacks. It can serve as an effective identifier for synthesized voices and holds potential for broader applications in audio copyright protection. Moreover, this framework boasts high flexibility, allowing for the combination of multiple watermark segments to achieve heightened robustness and expanded capacity. Utilizing 10 to 20-second audio as the host, our approach demonstrates an average Bit Error Rate (BER) of 0.48\% across ten common attacks, a remarkable reduction of over 2800\% in BER compared to the state-of-the-art watermarking tool. See https://aka.ms/wavmark for demos of our work.
Generating realistic audio effects for movies and other media is a challenging task that is accomplished today primarily through physical techniques known as Foley art. Foley artists create sounds with common objects (e.g., boxing gloves, broken glass) in time with video as it is playing to generate captivating audio tracks. In this work, we aim to develop a deep-learning based framework that does much the same - observes video in it's natural sequence and generates realistic audio to accompany it. Notably, we have reason to believe this is achievable due to advancements in realistic audio generation techniques conditioned on other inputs (e.g., Wavenet conditioned on text). We explore several different model architectures to accomplish this task that process both previously-generated audio and video context. These include deep-fusion CNN, dilated Wavenet CNN with visual context, and transformer-based architectures. We find that the transformer-based architecture yields the most promising results, matching low-frequencies to visual patterns effectively, but failing to generate more nuanced waveforms.
Text-based audio generation models have limitations as they cannot encompass all the information in audio, leading to restricted controllability when relying solely on text. To address this issue, we propose a novel model that enhances the controllability of existing pre-trained text-to-audio models by incorporating additional conditions including content (timestamp) and style (pitch contour and energy contour) as supplements to the text. This approach achieves fine-grained control over the temporal order, pitch, and energy of generated audio. To preserve the diversity of generation, we employ a trainable control condition encoder that is enhanced by a large language model and a trainable Fusion-Net to encode and fuse the additional conditions while keeping the weights of the pre-trained text-to-audio model frozen. Due to the lack of suitable datasets and evaluation metrics, we consolidate existing datasets into a new dataset comprising the audio and corresponding conditions and use a series of evaluation metrics to evaluate the controllability performance. Experimental results demonstrate that our model successfully achieves fine-grained control to accomplish controllable audio generation. Audio samples and our dataset are publicly available at https://conditionaudiogen.github.io/conditionaudiogen/
Building artificial intelligence (AI) systems on top of a set of foundation models (FMs) is becoming a new paradigm in AI research. Their representative and generative abilities learnt from vast amounts of data can be easily adapted and transferred to a wide range of downstream tasks without extra training from scratch. However, leveraging FMs in cross-modal generation remains under-researched when audio modality is involved. On the other hand, automatically generating semantically-relevant sound from visual input is an important problem in cross-modal generation studies. To solve this vision-to-audio (V2A) generation problem, existing methods tend to design and build complex systems from scratch using modestly sized datasets. In this paper, we propose a lightweight solution to this problem by leveraging foundation models, specifically CLIP, CLAP, and AudioLDM. We first investigate the domain gap between the latent space of the visual CLIP and the auditory CLAP models. Then we propose a simple yet effective mapper mechanism (V2A-Mapper) to bridge the domain gap by translating the visual input between CLIP and CLAP spaces. Conditioned on the translated CLAP embedding, pretrained audio generative FM AudioLDM is adopted to produce high-fidelity and visually-aligned sound. Compared to previous approaches, our method only requires a quick training of the V2A-Mapper. We further analyze and conduct extensive experiments on the choice of the V2A-Mapper and show that a generative mapper is better at fidelity and variability (FD) while a regression mapper is slightly better at relevance (CS). Both objective and subjective evaluation on two V2A datasets demonstrate the superiority of our proposed method compared to current state-of-the-art approaches - trained with 86% fewer parameters but achieving 53% and 19% improvement in FD and CS, respectively.
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 framework that utilizes the same learning method for speech, music, and sound effect generation. Our framework introduces 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 any modalities into LOA by using a GPT-2 model, and we perform self-supervised audio generation learning with a latent diffusion model conditioned on LOA. The proposed framework naturally brings advantages such as in-context learning abilities and reusable self-supervised pretrained AudioMAE and latent diffusion models. Experiments on the major benchmarks of text-to-audio, text-to-music, and text-to-speech demonstrate state-of-the-art or competitive performance against previous approaches. Our code, pretrained model, and demo are available at https://audioldm.github.io/audioldm2.