English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

Zero-shot text-to-speech (TTS) aims to synthesize voices with unseen speechprompts, which significantly reduces the data and computation requirements forvoice cloning by skipping the fine-tuning process. However, the promptingmechanisms of zero-shot TTS still face challenges in the following aspects: 1)previous works of zero-shot TTS are typically trained with single-sentenceprompts, which significantly restricts their performance when the data isrelatively sufficient during the inference stage. 2) The prosodic informationin prompts is highly coupled with timbre, making it untransferable to eachother. This paper introduces Mega-TTS 2, a generic prompting mechanism forzero-shot TTS, to tackle the aforementioned challenges. Specifically, we designa powerful acoustic autoencoder that separately encodes the prosody and timbreinformation into the compressed latent space while providing high-qualityreconstructions. Then, we propose a multi-reference timbre encoder and aprosody latent language model (P-LLM) to extract useful information frommulti-sentence prompts. We further leverage the probabilities derived frommultiple P-LLM outputs to produce transferable and controllable prosody.Experimental results demonstrate that Mega-TTS 2 could not only synthesizeidentity-preserving speech with a short prompt of an unseen speaker fromarbitrary sources but consistently outperform the fine-tuning method when thevolume of data ranges from 10 seconds to 5 minutes. Furthermore, our methodenables to transfer various speaking styles to the target timbre in afine-grained and controlled manner. Audio samples can be found inhttps://boostprompt.github.io/boostprompt/.

Mega-TTS 2: Boosting Prompting Mechanisms for Zero-Shot Speech Synthesis