initial commit

Cargo.lock

@@ -1834,13 +1834,14 @@ name = "sbv2_bindings"
 version = "0.1.0"
 dependencies = [
  "anyhow",
+ "ndarray",
  "pyo3",
  "sbv2_core",
 ]
 
 [[package]]
 name = "sbv2_core"
-version = "0.1.1"
+version = "0.1.2"
 dependencies = [
  "anyhow",
  "dotenvy",

Cargo.toml

@@ -6,3 +6,4 @@ members = ["sbv2_api", "sbv2_core", "sbv2_bindings"]
 anyhow = "1.0.86"
 dotenvy = "0.15.7"
 env_logger = "0.11.5"
+ndarray = "0.16.1"

sbv2_bindings/Cargo.toml

@@ -10,5 +10,6 @@ crate-type = ["cdylib"]
 
 [dependencies]
 anyhow.workspace = true
+ndarray.workspace = true
 pyo3 = { version = "0.22.0", features = ["anyhow"] }
 sbv2_core = { version = "0.1.0", path = "../sbv2_core" }

sbv2_bindings/examples/basic.py

@@ -0,0 +1,21 @@
+from sbv2_bindings import TTSModel
+
+
+def main():
+    with open("../models/debert.onnx", "rb") as f:
+        bert = f.read()
+    with open("../models/tokenizer.json", "rb") as f:
+        tokenizer = f.read()
+    
+    model = TTSModel(bert, tokenizer)
+
+    with open("../models/amitaro.sbv2", "rb") as f:
+        model.load_sbv2file(f.read())
+    
+    style_vector = model.get_style_vector("amitaro", 0, 1.0)
+    with open("output.wav", "wb") as f:
+        f.write(model.synthesize("こんにちは", "amitaro", style_vector, 0.0, 0.5))
+
+
+if __name__ == "__main__":
+    main()

sbv2_bindings/src/lib.rs

@@ -1,5 +1,6 @@
 use pyo3::prelude::*;
 mod sbv2;
+pub mod style;
 
 /// Formats the sum of two numbers as string.
 #[pyfunction]

sbv2_bindings/src/sbv2.rs

@@ -1,6 +1,8 @@
 use pyo3::prelude::*;
 use sbv2_core::tts::TTSModelHolder;
 
+use crate::style::StyleVector;
+
 #[pyclass]
 pub struct TTSModel {
     pub model: TTSModelHolder,
@@ -15,5 +17,40 @@ impl TTSModel {
         })
     }
 
-    fn load()
+    fn load_sbv2file(&mut self, ident: String, sbv2file_bytes: Vec<u8>) -> anyhow::Result<()> {
+        self.model.load_sbv2file(ident, sbv2file_bytes)?;
+        Ok(())
+    }
+
+    fn get_style_vector(
+        &self,
+        ident: String,
+        style_id: i32,
+        weight: f32,
+    ) -> anyhow::Result<StyleVector> {
+        Ok(StyleVector::new(
+            self.model.get_style_vector(ident, style_id, weight)?,
+        ))
+    }
+
+    fn synthesize(
+        &self,
+        text: String,
+        ident: String,
+        style_vector: StyleVector,
+        sdp_ratio: f32,
+        length_scale: f32,
+    ) -> anyhow::Result<Vec<u8>> {
+        let (bert_ori, phones, tones, lang_ids) = self.model.parse_text(&text)?;
+        Ok(self.model.synthesize(
+            ident,
+            bert_ori,
+            phones,
+            tones,
+            lang_ids,
+            style_vector.get(),
+            sdp_ratio,
+            length_scale,
+        )?)
+    }
 }

sbv2_bindings/src/style.rs

@@ -0,0 +1,16 @@
+use ndarray::Array1;
+use pyo3::prelude::*;
+
+#[pyclass]
+#[derive(Clone)]
+pub struct StyleVector(Array1<f32>);
+
+impl StyleVector {
+    pub fn new(data: Array1<f32>) -> Self {
+        StyleVector(data)
+    }
+
+    pub fn get(&self) -> Array1<f32> {
+        self.0.clone()
+    }
+}

sbv2_core/Cargo.toml

@@ -14,7 +14,7 @@ dotenvy.workspace = true
 env_logger.workspace = true
 hound = "3.5.1"
 jpreprocess = { version = "0.10.0", features = ["naist-jdic"] }
-ndarray = "0.16.1"
+ndarray.workspace = true
 num_cpus = "1.16.0"
 once_cell = "1.19.0"
 ort = { git = "https://github.com/pykeio/ort.git", version = "2.0.0-rc.6" }