DurIAN is a Duration Informed Attention Network for multimodal synthesis. This work focuses on an AI model architecture designed to generate coherent and synchronized outputs across multiple modalities.
The JSUT corpus is a free, large-scale Japanese speech dataset designed for end-to-end speech synthesis research. It provides valuable resources for developing advanced AI models in speech technology for the Japanese language.
Kitten TTS is a lightweight text-to-speech model. It features a live graphical user interface.
VITA-QinYu is the first expressive end-to-end (E2E) spoken language model that supports both role-playing and singing generation. It adopts a hybrid speech-text paradigm with multi-codebook audio tokens and was trained on 15.8K hours of data, outperforming other SLMs in expressiveness.
This paper proposes novel techniques for fine-grained speaking style control in prompt-based text-to-speech (TTS) models. It addresses inter-utterance style interpolation and intra-utterance style transitions, overcoming limitations of global style application.
This research addresses the Stability-Expressivity Gap in Spoken Language Models (SLMs) for low-resource languages, caused by the extensive use of synthetic data. While synthetic data improves phonetic accuracy, it degrades prosodic expressivity, a phenomenon termed Synthetic Erosion. The paper introduces self-alignment frameworks to recover expressivity.
This content introduces Aisha AI, Central Asia's fastest-growing AI platform, specializing in Uzbek-language speech synthesis, speech-to-text, chatbots, and voice agents. The guide provides 100 official links covering products, documentation, and applications across various sectors, driving digital transformation in the region.