diff --git a/paddlespeech/t2s/exps/fastspeech2/preprocess.py b/paddlespeech/t2s/exps/fastspeech2/preprocess.py
index db1842b2e89fe3044e96ca4babb07c1796d06da3..5fc51365a9b08ed56f5e89126a0ef54be71ecdd3 100644
--- a/paddlespeech/t2s/exps/fastspeech2/preprocess.py
+++ b/paddlespeech/t2s/exps/fastspeech2/preprocess.py
@@ -55,8 +55,11 @@ def process_sentence(config: Dict[str, Any],
     if utt_id in sentences:
         # reading, resampling may occur
         wav, _ = librosa.load(str(fp), sr=config.fs)
-        if len(wav.shape) != 1 or np.abs(wav).max() > 1.0:
+        if len(wav.shape) != 1:
             return record
+        max_value = np.abs(wav).max()
+        if max_value > 1.0:
+            wav = wav / max_value
         assert len(wav.shape) == 1, f"{utt_id} is not a mono-channel audio."
         assert np.abs(wav).max(
         ) <= 1.0, f"{utt_id} is seems to be different that 16 bit PCM."
diff --git a/paddlespeech/t2s/exps/gan_vocoder/preprocess.py b/paddlespeech/t2s/exps/gan_vocoder/preprocess.py
index 4871bca7130c418444152561975dff19bf629792..8adab0feaca5003142b81c396e340577a72ff549 100644
--- a/paddlespeech/t2s/exps/gan_vocoder/preprocess.py
+++ b/paddlespeech/t2s/exps/gan_vocoder/preprocess.py
@@ -47,8 +47,11 @@ def process_sentence(config: Dict[str, Any],
     if utt_id in sentences:
         # reading, resampling may occur
         y, _ = librosa.load(str(fp), sr=config.fs)
-        if len(y.shape) != 1 or np.abs(y).max() > 1.0:
+        if len(y.shape) != 1:
             return record
+        max_value = np.abs(y).max()
+        if max_value > 1.0:
+            y = y / max_value
         assert len(y.shape) == 1, f"{utt_id} is not a mono-channel audio."
         assert np.abs(y).max(
         ) <= 1.0, f"{utt_id} is seems to be different that 16 bit PCM."
diff --git a/paddlespeech/t2s/exps/speedyspeech/preprocess.py b/paddlespeech/t2s/exps/speedyspeech/preprocess.py
index e833d13940530f293842a842b65f33cf6d03d9bd..6c6b443fbaf970e5ca367eab4dc4ef413759c4df 100644
--- a/paddlespeech/t2s/exps/speedyspeech/preprocess.py
+++ b/paddlespeech/t2s/exps/speedyspeech/preprocess.py
@@ -47,8 +47,11 @@ def process_sentence(config: Dict[str, Any],
     if utt_id in sentences:
         # reading, resampling may occur
         wav, _ = librosa.load(str(fp), sr=config.fs)
-        if len(wav.shape) != 1 or np.abs(wav).max() > 1.0:
+        if len(wav.shape) != 1:
             return record
+        max_value = np.abs(wav).max()
+        if max_value > 1.0:
+            wav = wav / max_value
         assert len(wav.shape) == 1, f"{utt_id} is not a mono-channel audio."
         assert np.abs(wav).max(
         ) <= 1.0, f"{utt_id} is seems to be different that 16 bit PCM."
diff --git a/paddlespeech/t2s/exps/tacotron2/preprocess.py b/paddlespeech/t2s/exps/tacotron2/preprocess.py
index 14a0d7eae227f5650a716bc656f3d0c32ee077e3..95349d595b9154f1eb5f6ec68481fc50962e7117 100644
--- a/paddlespeech/t2s/exps/tacotron2/preprocess.py
+++ b/paddlespeech/t2s/exps/tacotron2/preprocess.py
@@ -51,8 +51,11 @@ def process_sentence(config: Dict[str, Any],
     if utt_id in sentences:
         # reading, resampling may occur
         wav, _ = librosa.load(str(fp), sr=config.fs)
-        if len(wav.shape) != 1 or np.abs(wav).max() > 1.0:
+        if len(wav.shape) != 1:
             return record
+        max_value = np.abs(wav).max()
+        if max_value > 1.0:
+            wav = wav / max_value
         assert len(wav.shape) == 1, f"{utt_id} is not a mono-channel audio."
         assert np.abs(wav).max(
         ) <= 1.0, f"{utt_id} is seems to be different that 16 bit PCM."