Large language models (LLMs) have revolutionized natural language processing (NLP) with impressive performance across various text-based tasks. However, the extension of text-dominant LLMs to with speech generation tasks remains underexplored. In this work, we introduce a text-to-speech (TTS) system powered by a fine-tuned Llama model, named TTS-Llama, that achieves state-of-the-art speech synthesis performance. Building on TTS-Llama, we further propose MoLE-Llama, a text-and-speech multimodal LLM developed through purely late-fusion parameter-efficient fine-tuning (PEFT) and a mixture-of-expert architecture. Extensive empirical results demonstrate MoLE-Llama’s competitive performance on both text-only question-answering (QA) and TTS tasks, mitigating catastrophic forgetting issue in either modality. Finally, we further explore MoLE-Llama in text-in-speech-out QA tasks, demonstrating its great potential as a multimodal dialog system capable of speech generation.

A singing voice conversion model converts a song in the voice of an arbitrary source singer to the voice of a target singer. Recently, methods that leverage self-supervised audio representations such as HuBERT and Wav2Vec 2.0 have helped further the state-of-the-art. Though these methods produce more natural and melodic singing outputs, they often rely on confusion and disentanglement losses to render the self-supervised representations speaker and pitch-invariant. In this paper, we circumvent disentanglement training and propose a new model that leverages ASR fine-tuned self-supervised representations as inputs to a HiFi-GAN neural vocoder for singing voice conversion. We experiment with different f0encoding schemes and show that an f0harmonic generation module that uses a parallel bank of transposed convolutions (PBTC) alongside ASR fine-tuned Wav2Vec 2.0 features results in the best singing voice conversion quality. Additionally, the model is capable of making a spoken voice sing. We also show that a simple f0shifting scheme during inference helps retain singer identity and bolsters the performance of our singing voice conversion model. Our results are backed up by extensive MOS studies that compare different ablations and baselines.

Most people who have tried to learn a foreign language would have experienced difficulties understanding or speaking with a native speaker’s accent. For native speakers, understanding or speaking a new accent is likewise a difficult task. An accent conversion system that changes a speaker’s accent but preserves that speaker’s voice identity, such as timbre and pitch, has the potential for a range of applications, such as communication, language learning, and entertainment. Existing accent conversion models tend to change the speaker identity and accent at the same time. Here, we use adversarial learning to disentangle accent dependent features while retaining other acoustic characteristics. What sets our work apart from existing accent conversion models is the capability to convert an unseen speaker’s utterance to multiple accents while preserving its original voice identity. Subjective evaluations show that our model generates audio that sound closer to the target accent and like the original speaker.

Neural vocoders, used for converting the spectral representations of an audio signal to the waveforms, are a commonly used component in speech synthesis pipelines. It focuses on synthesizing waveforms from low-dimensional representation, such as Mel-Spectrograms. In recent years, different approaches have been introduced to develop such vocoders. However, it becomes more challenging to assess these new vocoders and compare their performance to previous ones. To address this problem, we present VocBench, a framework that benchmark the performance of state-of-the-art neural vocoders. VocBench uses a systematic study to evaluate different neural vocoders in a shared environment that enables a fair comparison between them. In our experiments, we use the same setup for datasets, training pipeline, and evaluation metrics for all neural vocoders. We perform a subjective and objective evaluation to compare the performance of each vocoder along a different axis. Our results demonstrate that the framework can show competitive efficacy and quality of the synthesized samples for each vocoder. VocBench framework is available at https://github.com/facebookresearch/vocoder-benchmark.

Low bitrate speech codecs have become an area of intense research. Traditional speech codecs, which use signal processing methods to encode and decode speech, often suffer from quality issues at low bitrates. A neural speech codec, which uses a deep neural network in the compression pipeline, can help alleviate this issue. In this paper we present a new neural speech codec that: 1) supports variable bitrates 2) supports packet losses of up to 120 ms and 3) can operate at low-compute and high-compute modes. Our codec uses a hierarchical VQ-VAE (HVQVAE) for encoding and decoding spectral features at different bitrates. The decoded features are fed to a vocoder for speech synthesis. Depending upon the end user’s computing resources, the decoder either uses a powerful WaveRNN or a parametric vocoder for speech synthesis. Our experiments demonstrate that our HVQVAE + WaveRNN setup achieves high audio quality.

State of the art text-to-speech (TTS) models can generate high fidelity monolingual speech, but it is still challenging to synthesize multilingual speech from the same speaker. One major hurdle is for training data. It’s hard to find speakers who have native proficiency in several languages. One way of mitigating this issue is by generating polyglot corpus through voice conversion. In this paper, we train such multilingual TTS system through a novel cross-lingual voice conversion model trained with speaker-invariant features extracted from a speech representation model which is pre-trained with 53 languages through self-supervised learning [1]. To further improve the speaker identity shift, we also adopt a speaker similarity loss term during training. We then use this model to convert multilingual multi-speaker speech data to the voice of the target speaker. Through augmenting data from 4 other languages, we train a multilingual TTS system for a native monolingual English speaker which speaks 5 languages(English, French, German, Italian and Spanish). Our system achieves improved mean opinion score (MOS) compared with the baseline of multi-speaker system for all languages, specifically: 3.74 vs 3.62 for Spanish, 3.11 vs 2.71 for German, 3.47 vs 2.84 for Italian, and 2.72 vs 2.41 for French.

Typical high quality text-to-speech (TTS) systems today use a two-stage architecture, with a spectrum model stage that generates spectral frames and a vocoder stage that generates the actual audio. High-quality spectrum models usually incorporate the encoder-decoder architecture with self-attention or bi-directional long short-term (BLSTM) units. While these models can produce high quality speech, they often incur O(L) increase in both latency and real-time factor (RTF) with respect to input length L. In other words, longer inputs leads to longer delay and slower synthesis speed, limiting its use in real-time applications. In this paper, we propose a multi-rate attention architecture that breaks the latency and RTF bottlenecks by computing a compact representation during encoding and recurrently generating the attention vector in a streaming manner during decoding. The proposed architecture achieves high audio quality (MOS of 4.31 compared to groundtruth 4.48), low latency, and low RTF at the same time. Meanwhile, both latency and RTF of the proposed system stay constant regardless of input lengths, making it ideal for real-time applications.