#463·Amphion

[Docs]: Where can we find the resources for the paper: Advancing Zero-shot Text-to-Speech Intelligibility across Diverse Domains via Preference Alignment

Author: qhe-aiCreated Oct 14, 2025Updated Jan 17, 2026
Labelsdocumentation

Hi authors, thanks for the nice codebase! That's really impressive.

Documentation Reference

(Path/Link to the documentation file)

https://arxiv.org/pdf/2505.04113

Feedback on documentation

(Your suggestions to the documentation. e.g., accuracy, complex explanations, structural organization, practical examples, technical reliability, and consistency)

We open-source all resources used in this study at Amphion (Zhang et al., 2024c), including: (1) the proposed INTP and DPO-based alignment codebase for various TTS models, (2) all the INTP-enhanced models3 based on Ints, CosyVoice 2, ARS, F5-TTS, and MaskGCT, and (3) our newly constructed zero-shot TTS evaluation sets across diverse domains

According to the paper, the authors of the paper said they released their code in this repository. However, search the codebase and didn't find the related code. Where can I find the relevant code?

Many thanks!

Additional context

(Add any other context or screenshots about the documentation here.)