Search Results for "learning language"

...A key innovation is its use of continuous acoustic and semantic speech tokenizers operating at an ultra-low frame rate of 7.5 Hz, enabling high audio fidelity with efficient processing of long sequences. The model integrates a Qwen2.5-based large language model with a diffusion head to produce realistic acoustic details and capture conversational context. Training involved curriculum learning with increasing sequence lengths up to 65K tokens, allowing VibeVoice to handle very long dialogues effectively. Safety mechanisms include an audible disclaimer and imperceptible watermarking in all generated audio to mitigate misuse risks.

...The system introduces a multi-reward reinforcement learning framework that jointly optimizes for voice similarity, emotional expressiveness, pronunciation, and intelligibility, yielding output that can rival commercial options in naturalness and expressiveness. GLM-TTS also supports phoneme-level control and hybrid text + phoneme input, giving developers precise control over pronunciation critical for multilingual or polyphone­-rich languages.

fairseq2 is a modern, modular sequence modeling framework developed by Meta AI Research as a complete redesign of the original fairseq library. Built from the ground up for scalability, composability, and research flexibility, fairseq2 supports a broad range of language, speech, and multimodal content generation tasks, including instruction fine-tuning, reinforcement learning from human feedback (RLHF), and large-scale multilingual modeling. Unlike the original fairseq—which evolved into a large, monolithic codebase—fairseq2 introduces a clean, plugin-oriented architecture designed for long-term maintainability and rapid experimentation. ...

Step-Audio-EditX is an open-source, 3 billion-parameter audio model from StepFun AI designed to make expressive and precise editing of speech and audio as easy as text editing. Rather than treating audio editing as low-level waveform manipulation, this model converts speech into a sequence of discrete “audio tokens” (via a dual-codebook tokenizer) — combining a linguistic token stream and a semantic (prosody/emotion/style) token stream — thereby abstracting audio editing into high-level...

vits_chinese is an implementation of the VITS end-to-end text-to-speech (TTS) architecture tailored for Chinese (and possibly multilingual) speech synthesis. VITS is a model combining variational autoencoders (VAEs), normalizing flows, adversarial learning, and a stochastic duration predictor — a design that enables generation of natural, expressive speech, capturing variations in rhythm and prosody. By customizing or porting VITS for Chinese, this project aims to produce high-quality TTS outputs in a language that can be challenging due to tones, pronunciation variability, and prosody. ...

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. ...