Slava Shechtman

Avihu Dekel, Slava Shechtman, Raul Fernandez et al.

Large Language Models (LLMs) demonstrate impressive capabilities, yet interaction with these models is mostly facilitated through text. Using Text-To-Speech to synthesize LLM outputs typically results in notable latency, which is impractical for fluent voice conversations. We propose LLM2Speech, an architecture to synthesize speech while text is being generated by an LLM which yields significant latency reduction. LLM2Speech mimics the predictions of a non-streaming teacher model while limiting the exposure to future context in order to enable streaming. It exploits the hidden embeddings of the LLM, a by-product of the text generation that contains informative semantic context. Experimental results show that LLM2Speech maintains the teacher's quality while reducing the latency to enable natural conversations.

Vishal Sunder, Brian Kingsbury, George Saon et al.

In this paper, we take a step towards jointly modeling automatic speech recognition (STT) and speech synthesis (TTS) in a fully non-autoregressive way. We develop a novel multimodal framework capable of handling the speech and text modalities as input either individually or together. The proposed model can also be trained with unpaired speech or text data owing to its multimodal nature. We further propose an iterative refinement strategy to improve the STT and TTS performance of our model such that the partial hypothesis at the output can be fed back to the input of our model, thus iteratively improving both STT and TTS predictions. We show that our joint model can effectively perform both STT and TTS tasks, outperforming the STT-specific baseline in all tasks and performing competitively with the TTS-specific baseline across a wide range of evaluation metrics.

Kaizhi Qian, Xulin Fan, Junrui Ni et al.

Speech language models refer to language models with speech processing and understanding capabilities. One key desirable capability for speech language models is the ability to capture the intricate interdependency between content and prosody. The existing mainstream paradigm of training speech language models, which converts speech into discrete tokens before feeding them into LLMs, is sub-optimal in learning prosody information -- we find that the resulting LLMs do not exhibit obvious emerging prosody processing capabilities via pre-training alone. To overcome this, we propose ProsodyLM, which introduces a simple tokenization scheme amenable to learning prosody. Each speech utterance is first transcribed into text, followed by a sequence of word-level prosody tokens. Compared with conventional speech tokenization schemes, the proposed tokenization scheme retains more complete prosody information, and is more understandable to text-based LLMs. We find that ProsodyLM can learn surprisingly diverse emerging prosody processing capabilities through pre-training alone, ranging from harnessing the prosody nuances in generated speech, such as contrastive focus, understanding emotion and stress in an utterance, to maintaining prosody consistency in long contexts.

Slava Shechtman, Alex Sorin

Modern sequence to sequence neural TTS systems provide close to natural speech quality. Such systems usually comprise a network converting linguistic/phonetic features sequence to an acoustic features sequence, cascaded with a neural vocoder. The generated speech prosody (i.e. phoneme durations, pitch and loudness) is implicitly present in the acoustic features, being mixed with spectral information. Although the speech sounds natural, its prosody realization is randomly chosen and cannot be easily altered. The prosody control becomes an even more difficult task if no prosodic labeling is present in the training data. Recently, much progress has been achieved in unsupervised speaking style learning and generation, however human inspection is still required after the training for discovery and interpretation of the speaking styles learned by the system. In this work we introduce a fully automatic method that makes the system aware of the prosody and enables sentence-wise speaking pace and expressiveness control on a continuous scale. While being useful by itself in many applications, the proposed prosody control can also improve the overall quality and expressiveness of the synthesized speech, as demonstrated by subjective listening evaluations. We also propose a novel augmented attention mechanism, that facilitates better pace control sensitivity and faster attention convergence.

Zvi Kons, Slava Shechtman, Alex Sorin et al.

We present a lightweight adaptable neural TTS system with high quality output. The system is composed of three separate neural network blocks: prosody prediction, acoustic feature prediction and Linear Prediction Coding Net as a neural vocoder. This system can synthesize speech with close to natural quality while running 3 times faster than real-time on a standard CPU. The modular setup of the system allows for simple adaptation to new voices with a small amount of data. We first demonstrate the ability of the system to produce high quality speech when trained on large, high quality datasets. Following that, we demonstrate its adaptability by mimicking unseen voices using 5 to 20 minutes long datasets with lower recording quality. Large scale Mean Opinion Score quality and similarity tests are presented, showing that the system can adapt to unseen voices with quality gap of 0.12 and similarity gap of 3% compared to natural speech for male voices and quality gap of 0.35 and similarity of gap of 9 % for female voices.