Chatterbox Multilingual

此 Soniqo 页面记录本地 speech-swift / speech-core 实现中的 Chatterbox Multilingual。Hugging Face 包链接放在集成说明之后。

先进入站内页面

首页卡片和文档菜单先指向这里；源模型和权重包链接仍在本页提供。

概览

模型Chatterbox Multilingual
用途Multilingual zero-shot voice-cloning TTS
后端MLX fp16 on Apple; LiteRT default-voice runtime in Speech Core
输出24 kHz mono waveform
语言23 languages
许可证MIT
状态MLX cloning runtime; LiteRT greedy/default-voice runtime for edge deployments
来源Resemble AI Chatterbox
Swift 产品ChatterboxTTS
CLI / 运行时Programmatic speech-swift runtime; LiteRT example CLI in speech-core/examples/litert

使用

下面的片段对应当前 speech-swift 仓库暴露的 API 或命令。

import ChatterboxTTS

let model = try await ChatterboxTTSModel.fromPretrained()
let pcm = try model.clone(
    referenceSamples: reference,
    sampleRate: 24_000,
    text: "The cloned voice now speaks a new sentence.",
    languageId: "en"
)

模型链接

实现说明