Neural Codec Language Models are Zero-Shot Text to Speech Synthesizers

We introduce a language modeling approach for text to speech synthesis (TTS). Specifically, we train a neural codec language model (called Vall-E) using discrete codes derived from an off-the-shelf neural audio codec model, and regard TTS as a conditional language modeling task rather than continuous signal regression as in previous work. During the pre-training stage, we scale up the TTS training data to 60K hours of English speech which is hundreds of times larger than existing systems. Vall-E emerges in-context learning capabilities and can be used to synthesize high-quality personalized speech with only a 3-second enrolled recording of an unseen speaker as an acoustic prompt. Experiment results show that Vall-E significantly outperforms the state-of-the-art zero-shot TTS system in terms of speech naturalness and speaker similarity. In addition, we find Vall-E could preserve the speaker's emotion and acoustic environment of the acoustic prompt in synthesis. See https://aka.ms/valle for demos of our work.

Fast and accurate factorized neural transducer for text adaption of end-to-end speech recognition models

Neural transducer is now the most popular end-to-end model for speech recognition, due to its naturally streaming ability. However, it is challenging to adapt it with text-only data. Factorized neural transducer (FNT) model was proposed to mitigate this problem. The improved adaptation ability of FNT on text-only adaptation data came at the cost of lowered accuracy compared to the standard neural transducer model. We propose several methods to improve the performance of the FNT model. They are: adding CTC criterion during training, adding KL divergence loss during adaptation, using a pre-trained language model to seed the vocabulary predictor, and an efficient adaptation approach by interpolating the vocabulary predictor with the n-gram language model. A combination of these approaches results in a relative word-error-rate reduction of 9.48\% from the standard FNT model. Furthermore, n-gram interpolation with the vocabulary predictor improves the adaptation speed hugely with satisfactory adaptation performance.

VATLM: Visual-Audio-Text Pre-Training with Unified Masked Prediction for Speech Representation Learning

Although speech is a simple and effective way for humans to communicate with the outside world, a more realistic speech interaction contains multimodal information, e.g., vision, text. How to design a unified framework to integrate different modal information and leverage different resources (e.g., visual-audio pairs, audio-text pairs, unlabeled speech, and unlabeled text) to facilitate speech representation learning was not well explored. In this paper, we propose a unified cross-modal representation learning framework VATLM (Visual-Audio-Text Language Model). The proposed VATLM employs a unified backbone network to model the modality-independent information and utilizes three simple modality-dependent modules to preprocess visual, speech, and text inputs. In order to integrate these three modalities into one shared semantic space, VATLM is optimized with a masked prediction task of unified tokens, given by our proposed unified tokenizer. We evaluate the pre-trained VATLM on audio-visual related downstream tasks, including audio-visual speech recognition (AVSR), visual speech recognition (VSR) tasks. Results show that the proposed VATLM outperforms previous the state-of-the-art models, such as audio-visual pre-trained AV-HuBERT model, and analysis also demonstrates that VATLM is capable of aligning different modalities into the same space. To facilitate future research, we release the code and pre-trained models at https://aka.ms/vatlm.

Simulating realistic speech overlaps improves multi-talker ASR

Multi-talker automatic speech recognition (ASR) has been studied to generate transcriptions of natural conversation including overlapping speech of multiple speakers. Due to the difficulty in acquiring real conversation data with high-quality human transcriptions, a na\"ive simulation of multi-talker speech by randomly mixing multiple utterances was conventionally used for model training. In this work, we propose an improved technique to simulate multi-talker overlapping speech with realistic speech overlaps, where an arbitrary pattern of speech overlaps is represented by a sequence of discrete tokens. With this representation, speech overlapping patterns can be learned from real conversations based on a statistical language model, such as N-gram, which can be then used to generate multi-talker speech for training. In our experiments, multi-talker ASR models trained with the proposed method show consistent improvement on the word error rates across multiple datasets.

LongFNT: Long-form Speech Recognition with Factorized Neural Transducer

Traditional automatic speech recognition~(ASR) systems usually focus on individual utterances, without considering long-form speech with useful historical information, which is more practical in real scenarios. Simply attending longer transcription history for a vanilla neural transducer model shows no much gain in our preliminary experiments, since the prediction network is not a pure language model. This motivates us to leverage the factorized neural transducer structure, containing a real language model, the vocabulary predictor. We propose the {LongFNT-Text} architecture, which fuses the sentence-level long-form features directly with the output of the vocabulary predictor and then embeds token-level long-form features inside the vocabulary predictor, with a pre-trained contextual encoder RoBERTa to further boost the performance. Moreover, we propose the {LongFNT} architecture by extending the long-form speech to the original speech input and achieve the best performance. The effectiveness of our LongFNT approach is validated on LibriSpeech and GigaSpeech corpora with 19% and 12% relative word error rate~(WER) reduction, respectively.

Self-supervised learning with bi-label masked speech prediction for streaming multi-talker speech recognition

Self-supervised learning (SSL), which utilizes the input data itself for representation learning, has achieved state-of-the-art results for various downstream speech tasks. However, most of the previous studies focused on offline single-talker applications, with limited investigations in multi-talker cases, especially for streaming scenarios. In this paper, we investigate SSL for streaming multi-talker speech recognition, which generates transcriptions of overlapping speakers in a streaming fashion. We first observe that conventional SSL techniques do not work well on this task due to the poor representation of overlapping speech. We then propose a novel SSL training objective, referred to as bi-label masked speech prediction, which explicitly preserves representations of all speakers in overlapping speech. We investigate various aspects of the proposed system including data configuration and quantizer selection. The proposed SSL setup achieves substantially better word error rates on the LibriSpeechMix dataset.

Speech separation with large-scale self-supervised learning

Self-supervised learning (SSL) methods such as WavLM have shown promising speech separation (SS) results in small-scale simulation-based experiments. In this work, we extend the exploration of the SSL-based SS by massively scaling up both the pre-training data (more than 300K hours) and fine-tuning data (10K hours). We also investigate various techniques to efficiently integrate the pre-trained model with the SS network under a limited computation budget, including a low frame rate SSL model training setup and a fine-tuning scheme using only the part of the pre-trained model. Compared with a supervised baseline and the WavLM-based SS model using feature embeddings obtained with the previously released 94K hours trained WavLM, our proposed model obtains 15.9% and 11.2% of relative word error rate (WER) reductions, respectively, for a simulated far-field speech mixture test set. For conversation transcription on real meeting recordings using continuous speech separation, the proposed model achieves 6.8% and 10.6% of relative WER reductions over the purely supervised baseline on AMI and ICSI evaluation sets, respectively, while reducing the computational cost by 38%.

Streaming, fast and accurate on-device Inverse Text Normalization for Automatic Speech Recognition

Automatic Speech Recognition (ASR) systems typically yield output in lexical form. However, humans prefer a written form output. To bridge this gap, ASR systems usually employ Inverse Text Normalization (ITN). In previous works, Weighted Finite State Transducers (WFST) have been employed to do ITN. WFSTs are nicely suited to this task but their size and run-time costs can make deployment on embedded applications challenging. In this paper, we describe the development of an on-device ITN system that is streaming, lightweight & accurate. At the core of our system is a streaming transformer tagger, that tags lexical tokens from ASR. The tag informs which ITN category might be applied, if at all. Following that, we apply an ITN-category-specific WFST, only on the tagged text, to reliably perform the ITN conversion. We show that the proposed ITN solution performs equivalent to strong baselines, while being significantly smaller in size and retaining customization capabilities.

LAMASSU: Streaming Language-Agnostic Multilingual Speech Recognition and Translation Using Neural Transducers

End-to-end formulation of automatic speech recognition (ASR) and speech translation (ST) makes it easy to use a single model for both multilingual ASR and many-to-many ST. In this paper, we propose streaming language-agnostic multilingual speech recognition and translation using neural transducers (LAMASSU). To enable multilingual text generation in LAMASSU, we conduct a systematic comparison between specified and unified prediction and joint networks. We leverage a language-agnostic multilingual encoder that substantially outperforms shared encoders. To enhance LAMASSU, we propose to feed target LID to encoders. We also apply connectionist temporal classification regularization to transducer training. Experimental results show that LAMASSU not only drastically reduces the model size but also outperforms monolingual ASR and bilingual ST models.

A Weakly-Supervised Streaming Multilingual Speech Model with Truly Zero-Shot Capability

In this paper, we introduce our work of building a Streaming Multilingual Speech Model (SM2), which can transcribe or translate multiple spoken languages into texts of the target language. The backbone of SM2 is Transformer Transducer, which has high streaming capability. Instead of human labeled speech translation (ST) data, SM2 models are trained using weakly supervised data generated by converting the transcriptions in speech recognition corpora with a machine translation service. With 351 thousand hours of anonymized speech training data from 25 languages, SM2 models achieve comparable or even better ST quality than some recent popular large-scale non-streaming speech models. More importantly, we show that SM2 has the truly zero-shot capability when expanding to new target languages, yielding high quality ST results for {source-speech, target-text} pairs that are not seen during training.