ggml-org/llama.cpp

tts : add support for Orpheus

Open

#12,476 创建于 2025年3月20日

在 GitHub 查看
 (7 评论) (26 反应) (0 负责人)C++ (110,169 star) (18,202 fork)batch import
good first issuetts

描述

HF: https://huggingface.co/collections/canopylabs/orpheus-tts-67d9ea3f6c05a941c06ad9d2

These TTS models seem suitable for supporting. To do that, we need to implement the SNAC audio codec: https://github.com/hubertsiuzdak/snac/

Sample implementation using Python-based inference of SNAC: https://github.com/isaiahbjork/orpheus-tts-local

Similar model support (OuteTTS): https://github.com/ggml-org/llama.cpp/pull/10784 Can be used as a reference how to implement this.

贡献者指南

技术栈
cpp
领域
machine learningai
议题类型
feature
难度面向新贡献者的预计实现难度，1 表示很小改动，5 表示专家级工作。
3
预计时间有经验贡献者完成调查、实现、测试并准备 pull request 的粗略时间范围。
1-2 days
活动状态议题当前的可参与程度：新鲜、活跃、陈旧、阻塞或等待维护者输入。
active
清晰度议题是否清楚说明期望改动、验收标准和下一步。
clear
前置要求
C++audio codec fundamentalsllama.cpp model integration
新手友好度1-100 的估计分数，表示该议题对首次贡献者的友好程度。
35
研究方向
Study the linked OuteTTS PR (https://github.com/ggml org/llama.cpp/pull/10784) to understand the pattern for adding new TTS model support. Then implement the SNAC audio codec as described in the SNAC repository (https://github.com/hubertsiuzdak/snac/) and adapt the Orpheus TTS model using the provided HF collection and sample implementation. The implementation will likely involve new model architecture files in the llama.cpp codebase.