Repository navigation
Releases
Releases · launcher-rs/audio-cpp-rs
v0.10.0
新增
- 跟进上游 audio.cpp(submodule 从
d2da530b前进到5e7273db),新增 5 个模型族(全量 loader 族 147 → 151,完整清单见 模型目录):m5_tts:M5-TTS 多语自回归 TTS + 参考音频克隆(Qwen3 speech codec 解码;zh/yue/zh-Sichuan/en/de/fr/ja/ko;离线+流式;featuremodel-m5-tts)。tiny_voca:Tiny Voca 4.5M 参数英文 TTS,自带 dizione 音素化前端、无需外部 phonemizer,内置 M2/F2 两音色,支持长文本(仅离线;featuremodel-tiny-voca)。rosvot:ROSVOT 歌声转符号谱——音符音高/时值与词对齐 MIDI 转写,输出 MIDI 或 JSON 音符事件(仅离线;featuremodel-rosvot)。soulx_singer:SoulX-Singer 参考音频条件的歌唱合成(SVS)与歌唱 voice conversion(SVC),zh/en/yue(仅离线;taskmusic/svc;featuremodel-soulx-singer)。granite_speech5:IBM Granite Speech 5.0 2B 自回归多语 ASR + 语音翻译,en/fr/de/es/pt/ja,翻译目标另支持 it/zh(仅离线;featuremodel-granite-speech5)。
ModelFamily新增对应枚举变体与from_path()识别,TaskKind无需变化(svc为既有任务类型)。
变更
- 上游模型行为随 submodule 升级带入:
zonos2新增长文本续写选项long_form_continuation且分块默认值 300→150;cosyvoice3新增speed语速选项;ace_step新增reference_audio音色条件;granite5asr编码器图 arena 默认值调整。全部经通用 options 字符串透传,audio-cpp的 serde 结构向后兼容。 - 推送
v*标签触发的预编译工作流现在会校验标签与Cargo.toml版本一致,并从本文件提取对应版本段作为 GitHub Release 说明(缺失时退回通用文案并告警)。
本 Release 附带预编译静态库:共 9 个资产(linux 4 / windows 4 / macos 1),文件名含 submodule commit(audio_commit),包内 metadata.json 记录 target/features/crate_version;audio-cpp-sys 的 prebuilt feature 按 crate 版本号自动下载对应 tag 的资产。
v0.9.0
v0.8.1
v0.8.0
v0.7.0
v0.6.0
v0.5.0
v0.4.0
v0.3.1
Prebuilt audio-cpp (engine_runtime) static libraries for audio-cpp-rs 0.3.1. Submodule audio.cpp @ da16c1b (0.7.0 dev line). cpu/vulkan/metal x full model set. Windows assets added by a follow-up build.