diff --git a/output.wav b/output.wav
index 7883ed9..0a301dd 100644
Binary files a/output.wav and b/output.wav differ
diff --git a/sbv2_core/src/jtalk.rs b/sbv2_core/src/jtalk.rs
index 201a651..c553abb 100644
--- a/sbv2_core/src/jtalk.rs
+++ b/sbv2_core/src/jtalk.rs
@@ -223,27 +223,6 @@ impl JTalkProcess {
     }
 
     fn kata_to_phoneme_list(mut text: String) -> Result<Vec<String>> {
-        /*
-        if set(text).issubset(set(PUNCTUATIONS)):
-            return list(text)
-        # `text` がカタカナ（`ー`含む）のみからなるかどうかをチェック
-        if __KATAKANA_PATTERN.fullmatch(text) is None:
-            raise ValueError(f"Input must be katakana only: {text}")
-
-        def mora2phonemes(mora: str) -> str:
-            consonant, vowel = MORA_KATA_TO_MORA_PHONEMES[mora]
-            if consonant is None:
-                return f" {vowel}"
-            return f" {consonant} {vowel}"
-
-        spaced_phonemes = __MORA_PATTERN.sub(lambda m: mora2phonemes(m.group()), text)
-
-        # 長音記号「ー」の処理
-        long_replacement = lambda m: m.group(1) + (" " + m.group(1)) * len(m.group(2))  # type: ignore
-        spaced_phonemes = __LONG_PATTERN.sub(long_replacement, spaced_phonemes)
-
-        return spaced_phonemes.strip().split(" ")
-        */
         if PUNCTUATIONS.contains(&text.as_str()) {
             return Ok(text.chars().map(|x| x.to_string()).collect());
         }
diff --git a/sbv2_core/src/main.rs b/sbv2_core/src/main.rs
index 63ad56b..b0eb401 100644
--- a/sbv2_core/src/main.rs
+++ b/sbv2_core/src/main.rs
@@ -1,7 +1,7 @@
 use sbv2_core::{error, tts};
 
 fn main() -> error::Result<()> {
-    let text = "おはようございます。";
+    let text = "眠たい";
 
     let tts_model = tts::TTSModel::new(
         "models/debert.onnx",
@@ -12,7 +12,9 @@ fn main() -> error::Result<()> {
     let (bert_ori, phones, tones, lang_ids) = tts_model.parse_text(text)?;
 
     let style_vector = tts_model.get_style_vector(0, 1.0)?;
-    tts_model.synthesize(bert_ori.to_owned(), phones, tones, lang_ids, style_vector)?;
+    let data = tts_model.synthesize(bert_ori.to_owned(), phones, tones, lang_ids, style_vector)?;
+
+    std::fs::write("output.wav", data)?;
 
     Ok(())
 }
diff --git a/sbv2_core/src/model.rs b/sbv2_core/src/model.rs
index d5c0eb2..ae6c535 100644
--- a/sbv2_core/src/model.rs
+++ b/sbv2_core/src/model.rs
@@ -1,7 +1,8 @@
 use crate::error::Result;
 use hound::{SampleFormat, WavSpec, WavWriter};
-use ndarray::{array, Array1, Array2, Axis};
+use ndarray::{array, s, Array1, Array2, Axis};
 use ort::{GraphOptimizationLevel, Session};
+use std::io::Cursor;
 
 pub fn load_model(model_file: &str) -> Result<Session> {
     let session = Session::builder()?
@@ -11,24 +12,6 @@ pub fn load_model(model_file: &str) -> Result<Session> {
     Ok(session)
 }
 
-fn write_wav(file_path: &str, audio: &[f32], sample_rate: u32) -> Result<()> {
-    let spec = WavSpec {
-        channels: 1, // モノラルの場合。ステレオなどの場合は2に変更
-        sample_rate,
-        bits_per_sample: 16,
-        sample_format: SampleFormat::Int,
-    };
-
-    let mut writer = WavWriter::create(file_path, spec)?;
-    for &sample in audio {
-        let int_sample = (sample * i16::MAX as f32).clamp(i16::MIN as f32, i16::MAX as f32) as i16;
-        writer.write_sample(int_sample)?;
-    }
-    writer.finalize()?;
-
-    Ok(())
-}
-
 pub fn synthesize(
     session: &Session,
     bert_ori: Array2<f32>,
@@ -36,7 +19,7 @@ pub fn synthesize(
     tones: Array1<i64>,
     lang_ids: Array1<i64>,
     style_vector: Array1<f32>,
-) -> Result<()> {
+) -> Result<Vec<u8>> {
     let bert = bert_ori.insert_axis(Axis(0));
     let x_tst_lengths: Array1<i64> = array![x_tst.shape()[0] as i64];
     let x_tst = x_tst.insert_axis(Axis(0));
@@ -53,7 +36,30 @@ pub fn synthesize(
         "ja_bert" => style_vector,
     }?)?;
 
-    let audio_array = outputs.get("output").unwrap().try_extract_tensor::<f32>()?;
-    write_wav("output.wav", audio_array.as_slice().unwrap(), 44100)?;
-    Ok(())
+    let audio_array = outputs
+        .get("output")
+        .unwrap()
+        .try_extract_tensor::<f32>()?
+        .to_owned();
+
+    let buffer = {
+        let spec = WavSpec {
+            channels: 1,
+            sample_rate: 44100,
+            bits_per_sample: 32,
+            sample_format: SampleFormat::Float,
+        };
+        let mut cursor = Cursor::new(Vec::new());
+        let mut writer = WavWriter::new(&mut cursor, spec)?;
+        for i in 0..audio_array.shape()[0] {
+            let output = audio_array.slice(s![i, 0, ..]).to_vec();
+            for sample in output {
+                writer.write_sample(sample)?;
+            }
+        }
+        writer.finalize()?;
+        cursor.into_inner()
+    };
+
+    Ok(buffer)
 }
diff --git a/sbv2_core/src/tts.rs b/sbv2_core/src/tts.rs
index 26b7434..a7537d8 100644
--- a/sbv2_core/src/tts.rs
+++ b/sbv2_core/src/tts.rs
@@ -107,8 +107,8 @@ impl TTSModel {
         tones: Array1<i64>,
         lang_ids: Array1<i64>,
         style_vector: Array1<f32>,
-    ) -> Result<()> {
-        model::synthesize(
+    ) -> Result<Vec<u8>> {
+        let buffer = model::synthesize(
             &self.vits2,
             bert_ori.to_owned(),
             phones,
@@ -116,6 +116,6 @@ impl TTSModel {
             lang_ids,
             style_vector,
         )?;
-        Ok(())
+        Ok(buffer)
     }
 }