diff --git a/musicagent/auxiliary/muzic/roc/midi2dict.py b/musicagent/auxiliary/muzic/roc/midi2dict.py
new file mode 100644
index 00000000..e827d4ab
--- /dev/null
+++ b/musicagent/auxiliary/muzic/roc/midi2dict.py
@@ -0,0 +1,253 @@
+import miditoolkit
+
+def midi_to_lyrics(midi) -> dict:
+    # Load MIDI file
+    # midi = miditoolkit.MidiFile(midi_path, charset='utf-8')
+
+    ticks_per_beat = midi.ticks_per_beat
+    tempo = midi.tempo_changes[0].tempo
+
+    def ticks_to_time(ticks):
+        return ticks / ticks_per_beat * 60 / tempo
+
+    # Extract note events and their timing information
+    note_events = []
+    for note in midi.instruments[0].notes:
+        note_events.append({
+            'start': ticks_to_time(note.start),
+            'end': ticks_to_time(note.end),
+            'pitch': pitch_to_note[note.pitch],
+            'velocity': note.velocity
+        })
+
+    # Sort note events by start time
+    note_events = sorted(note_events, key=lambda x: x['start'])
+
+    # get lyric events with start time
+    lyric_events = []
+    for lyric in midi.lyrics:
+        lyric_events.append({
+            'start': ticks_to_time(lyric.time),
+            'text': lyric.text
+        })
+    
+    # Sort lyrics by start time
+    lyric_events = sorted(lyric_events, key=lambda x: x['start'])
+
+    # Initialize output variables
+    notes = []
+    notes_duration = []
+    lyrics = []
+    processed = False
+
+    if note_events[0]['start'] > 0:
+        lyrics.append('AP')
+        notes.append(['rest'])
+        notes_duration.append([note_events[0]['start']])
+
+    def remove_nota(text):
+        exist = False
+        for notation in notations:
+            if notation in text:
+                text = text.replace(notation, '')
+                exist = True
+
+        return text, exist
+
+    # Loop through note events and generate lyrics and note information
+    while len(note_events) > 0 and len(lyric_events) > 0:
+        note_event = note_events[0]
+
+        # Convert note start and duration to time in seconds
+        start_time = note_event['start']
+        duration_time = note_event['end'] - note_event['start']
+
+        if len(lyric_events) > 1 and start_time >= lyric_events[1]['start']:
+            if bias > 0 and sep:
+                notes_duration[-1][-1] -= bias
+                lyrics.append('AP')
+                notes.append(['rest'])
+                notes_duration.append([bias])
+                sep = False
+
+            lyric_events.pop(0)
+            processed = False
+
+        if not processed:
+            text, sep = remove_nota(lyric_events[0]['text'])
+            lyrics.append(text)
+            notes.append([])
+            notes_duration.append([])
+            processed = True
+
+        notes[-1].append(note_event['pitch'])
+
+        if len(note_events) > 1:
+            duration_time = note_events[1]['start'] - note_event['start']
+            bias = note_events[1]['start'] - note_event['end']
+        else:
+            duration_time = note_event['end'] - note_event['start']
+            bias = 0
+        
+        notes_duration[-1].append(duration_time)
+
+        if duration_time == 0:
+            notes_duration[-1] = notes_duration[-1][:-1]
+            notes[-1] = notes[-1][:-1]
+
+        note_events.pop(0)
+
+    i = 0
+    while i < len(notes):
+        if len(notes[i]) == 0:
+            notes.pop(i)
+            notes_duration.pop(i)
+            lyrics.pop(i)
+        else:
+            i += 1
+
+    assert len(lyrics) == len(notes) 
+    assert len(lyrics) == len(notes_duration)
+    # Return output dictionary
+    output = {
+        'text': ''.join(lyrics),
+        'notes': ' | '.join([' '.join(note) for note in notes]),
+        'notes_duration': ' | '.join([' '.join([str(d) for d in duration]) for duration in notes_duration]),
+        'input_type': 'word'
+    }
+    print(sum(sum(notes_duration, [])))
+    # print(output)
+    return output
+
+
+notations = [',', '.', '，', '。', '!', '?', '！', '？']
+pitch_to_note = {
+0: 'C-1',
+1: 'C#-1/Db-1',
+2: 'D-1',
+3: 'D#-1/Eb-1',
+4: 'E-1',
+5: 'F-1',
+6: 'F#-1/Gb-1',
+7: 'G-1',
+8: 'G#-1/Ab-1',
+9: 'A-1',
+10: 'A#-1/Bb-1',
+11: 'B-1',
+12: 'C0',
+13: 'C#0/Db0',
+14: 'D0',
+15: 'D#0/Eb0',
+16: 'E0',
+17: 'F0',
+18: 'F#0/Gb0',
+19: 'G0',
+20: 'G#0/Ab0',
+21: 'A0',
+22: 'A#0/Bb0',
+23: 'B0',
+24: 'C1',
+25: 'C#1/Db1',
+26: 'D1',
+27: 'D#1/Eb1',
+28: 'E1',
+29: 'F1',
+30: 'F#1/Gb1',
+31: 'G1',
+32: 'G#1/Ab1',
+33: 'A1',
+34: 'A#1/Bb1',
+35: 'B1',
+36: 'C2',
+37: 'C#2/Db2',
+38: 'D2',
+39: 'D#2/Eb2',
+40: 'E2',
+41: 'F2',
+42: 'F#2/Gb2',
+43: 'G2',
+44: 'G#2/Ab2',
+45: 'A2',
+46: 'A#2/Bb2',
+47: 'B2',
+48: 'C3',
+49: 'C#3/Db3',
+50: 'D3',
+51: 'D#3/Eb3',
+52: 'E3',
+53: 'F3',
+54: 'F#3/Gb3',
+55: 'G3',
+56: 'G#3/Ab3',
+57: 'A3',
+58: 'A#3/Bb3',
+59: 'B3',
+60: 'C4',
+61: 'C#4/Db4',
+62: 'D4',
+63: 'D#4/Eb4',
+64: 'E4',
+65: 'F4',
+66: 'F#4/Gb4',
+67: 'G4',
+68: 'G#4/Ab4',
+69: 'A4',
+70: 'A#4/Bb4',
+71: 'B4',
+72: 'C5',
+73: 'C#5/Db5',
+74: 'D5',
+75: 'D#5/Eb5',
+76: 'E5',
+77: 'F5',
+78: 'F#5/Gb5',
+79: 'G5',
+80: 'G#5/Ab5',
+81: 'A5',
+82: 'A#5/Bb5',
+83: 'B5',
+84: 'C6',
+85: 'C#6/Db6',
+86: 'D6',
+87: 'D#6/Eb6',
+88: 'E6',
+89: 'F6',
+90: 'F#6/Gb6',
+91: 'G6',
+92: 'G#6/Ab6',
+93: 'A6',
+94: 'A#6/Bb6',
+95: 'B6',
+96: 'C7',
+97: 'C#7/Db7',
+98: 'D7',
+99: 'D#7/Eb7',
+100: 'E7',
+101: 'F7',
+102: 'F#7/Gb7',
+103: 'G7',
+104: 'G#7/Ab7',
+105: 'A7',
+106: 'A#7/Bb7',
+107: 'B7',
+108: 'C8',
+109: 'C#8/Db8',
+110: 'D8',
+111: 'D#8/Eb8',
+112: 'E8',
+113: 'F8',
+114: 'F#8/Gb8',
+115: 'G8',
+116: 'G#8/Ab8',
+117: 'A8',
+118: 'A#8/Bb8',
+119: 'B8',
+120: 'C9',
+121: 'C#9/Db9',
+122: 'D9',
+123: 'D#9/Eb9',
+124: 'E9',
+125: 'F9',
+126: 'F#9/Gb9',
+127: 'G9'
+}