diff --git a/volia/data.py b/volia/data.py
index 744a03a..b7afb0b 100644
--- a/volia/data.py
+++ b/volia/data.py
@@ -8,7 +8,6 @@ import core.data
 
 
 
-
 def filter_file(file, filter, outfile):
     file_path = file
     filter_path = filter
@@ -24,8 +23,28 @@ def filter_file(file, filter, outfile):
     
 
 def convert(file, type_from, type):
+
     pass
 
+def utt2dur(wavscp: str, outfile: str):
+    
+    import wave
+    import contextlib
+
+    with open(wavscp, "r") as f, open(outfile, "w") as of:
+        for line in f:
+            splited = line.replace("\n", "").split(" ")
+            id_ = splited[0]
+            wav_ = splited[1]
+            duration = 0
+            
+            with contextlib.closing(wave.open(wav_,'r')) as wav_f:
+                frames = wav_f.getnframes()
+                rate = wav_f.getframerate()
+                duration = frames / float(rate)
+            
+            of.write(f"{id_} {duration}\n")
+
 
 if __name__ == "__main__":
     # Main parser
@@ -46,6 +65,16 @@ if __name__ == "__main__":
     parser_convert.add_argument("--type", choices=["old-masseffect", "new-masseffect"], required=True)
     parser_convert.set_defaults(which="convert")
 
+    # utt2dur
+    parser_utt2dur = subparsers.add_parser("utt2dur", help="generate utt2dur file")
+    parser_utt2dur.add_argument("--wavscp", type=str, help="wav file", required=True)
+    parser_utt2dur.add_argument("--outfile", type=str, default="utt2dur", help="output file")
+    parser_utt2dur.set_defaults(which="utt2dur")
+
+    # TODO: utt2label_to_label2utt
+
+    # TODO: label2utt_to_utt2label
+
     # Parse
     args = parser.parse_args()
 
@@ -53,6 +82,7 @@ if __name__ == "__main__":
     runner = SubCommandRunner({
         "convert" : convert,
         "filter": filter_file,
+        "utt2dur": utt2dur
     })
 
     runner.run(args.which, args.__dict__, remove="which")