|
|
--- |
|
|
license: cc-by-4.0 |
|
|
--- |
|
|
|
|
|
# DiaMoE-TTS: A Unified IPA-based Dialect TTS Framework with Mixture-of-Experts and Parameter-Efficient Zero-Shot Adaptation |
|
|
|
|
|
github:[DiaMoE-TTS](https://github.com/GiantAILab/DiaMoE-TTS) |
|
|
|
|
|
We utilize the [Common Voice Cantonese dataset](https://arxiv.org/abs/1912.06670), the [Emilia Mandarin dataset](https://arxiv.org/abs/2407.05361) and dialectal data from the [KeSpeech corpus](https://openreview.net/forum?id=b3Zoeq2sCLq) and a open-source [Sourthern Min dataset](https://sutian.moe.edu.tw/zh-hant/siongkuantsuguan/) for training. |
|
|
We only release the frontend of the open-source dataset IPA here, audio data that matches the IPA frontend can be applied for or downloaded from the corresponding official link. |
|
|
|
|
|
## Short Intro |
|
|
|
|
|
Dialect speech embodies rich cultural and linguistic diversity, yet building text-to-speech (TTS) systems for dialects remains challenging due to scarce data, inconsistent orthographies, and complex phonetic variation. To address these issues, we present DiaMoE-TTS, a unified IPA-based framework that standardizes phonetic representations and resolves grapheme-to-phoneme ambiguities. Built upon the F5-TTS architecture, the system introduces a dialect-aware Mixture-of-Experts (MoE) to model phonological differences and employs parameter-efficient adaptation with Low-Rank Adaptors (LoRA) and Conditioning Adapters for rapid transfer to new dialects. Unlike approaches dependent on large-scale or proprietary resources, DiaMoE-TTS enables scalable, open-data-driven synthesis. Experiments demonstrate natural and expressive speech generation, achieving zero-shot performance on unseen dialects and specialized domains such as Peking Opera with only a few hours of data. |
|
|
|
