jbonsai

日语

用于语音合成应用的文本到语音库。

"jbonsai" 将完整上下文标签序列转换为音频波形。

此项目目前是HTS Engine在Rust语言中的重写（这可能随时改变，并且不能保证jbonsai产生与HTS Engine相同的结果）。

目标

• 尽可能提高可读性。
• 在不影响可读性的前提下，
  • 提高速度。
  • 保持内存消耗低。
• 可以编译为WebAssembly。

用法

将以下内容放入 Cargo.toml。

[dependencies]
jbonsai = "0.1.0"

示例

此示例在 speech 变量中生成一个单声道、48,000 Hz（通常是）PCM数据，表示“盆栽”（ぼんさい; bonsai）。

# fn main() -> Result<(), Box<dyn std::error::Error>> {
// 盆栽,名詞,一般,*,*,*,*,盆栽,ボンサイ,ボンサイ,0/4,C2
let lines = [
    "xx^xx-sil+b=o/A:xx+xx+xx/B:xx-xx_xx/C:xx_xx+xx/D:xx+xx_xx/E:xx_xx!xx_xx-xx/F:xx_xx#xx_xx@xx_xx|xx_xx/G:4_4%0_xx_xx/H:xx_xx/I:xx-xx@xx+xx&xx-xx|xx+xx/J:1_4/K:1+1-4",
    "xx^sil-b+o=N/A:-3+1+4/B:xx-xx_xx/C:02_xx+xx/D:xx+xx_xx/E:xx_xx!xx_xx-xx/F:4_4#0_xx@1_1|1_4/G:xx_xx%xx_xx_xx/H:xx_xx/I:1-4@1+1&1-1|1+4/J:xx_xx/K:1+1-4",
    "sil^b-o+N=s/A:-3+1+4/B:xx-xx_xx/C:02_xx+xx/D:xx+xx_xx/E:xx_xx!xx_xx-xx/F:4_4#0_xx@1_1|1_4/G:xx_xx%xx_xx_xx/H:xx_xx/I:1-4@1+1&1-1|1+4/J:xx_xx/K:1+1-4",
    "b^o-N+s=a/A:-2+2+3/B:xx-xx_xx/C:02_xx+xx/D:xx+xx_xx/E:xx_xx!xx_xx-xx/F:4_4#0_xx@1_1|1_4/G:xx_xx%xx_xx_xx/H:xx_xx/I:1-4@1+1&1-1|1+4/J:xx_xx/K:1+1-4",
    "o^N-s+a=i/A:-1+3+2/B:xx-xx_xx/C:02_xx+xx/D:xx+xx_xx/E:xx_xx!xx_xx-xx/F:4_4#0_xx@1_1|1_4/G:xx_xx%xx_xx_xx/H:xx_xx/I:1-4@1+1&1-1|1+4/J:xx_xx/K:1+1-4",
    "N^s-a+i=sil/A:-1+3+2/B:xx-xx_xx/C:02_xx+xx/D:xx+xx_xx/E:xx_xx!xx_xx-xx/F:4_4#0_xx@1_1|1_4/G:xx_xx%xx_xx_xx/H:xx_xx/I:1-4@1+1&1-1|1+4/J:xx_xx/K:1+1-4",
    "s^a-i+sil=xx/A:0+4+1/B:xx-xx_xx/C:02_xx+xx/D:xx+xx_xx/E:xx_xx!xx_xx-xx/F:4_4#0_xx@1_1|1_4/G:xx_xx%xx_xx_xx/H:xx_xx/I:1-4@1+1&1-1|1+4/J:xx_xx/K:1+1-4",
    "a^i-sil+xx=xx/A:xx+xx+xx/B:xx-xx_xx/C:xx_xx+xx/D:xx+xx_xx/E:4_4!0_xx-xx/F:xx_xx#xx_xx@xx_xx|xx_xx/G:xx_xx%xx_xx_xx/H:1_4/I:xx-xx@xx+xx&xx-xx|xx+xx/J:xx_xx/K:1+1-4",
];
let engine = jbonsai::Engine::load(&[
    // The path to the `.htsvoice` model file.
    // Currently only Japanese models are supported (due to the limitation of jlabel).
    "models/hts_voice_nitech_jp_atr503_m001-1.05/nitech_jp_atr503_m001.htsvoice",
])?;
let speech = engine.synthesize(&lines)?;
println!(
    "The synthesized voice has {} samples in total.",
    speech.len()
);
# Ok(())
# }

版权

此软件包含以下源代码：

• hts_engine API.
  • 版权（c）2001-2014 爱知工业大学 计算机科学部
  • 版权（c）2001-2008 东京工业大学 理工研究生院

许可

BSD-3-Clause