diff --git a/lrcfetch/cli.py b/lrcfetch/cli.py
index 9f9ab8a..8d8a3e3 100644
--- a/lrcfetch/cli.py
+++ b/lrcfetch/cli.py
@@ -13,7 +13,7 @@ from lrcfetch.core import LrcManager
 
 app = typer.Typer(
     help="LRCFetch — Fetch line-synced lyrics for your music player.",
-    add_completion=False,
+    add_completion=True,
 )
 
 manager = LrcManager()
diff --git a/lrcfetch/lrc.py b/lrcfetch/lrc.py
index 578d4b4..5a9eff2 100644
--- a/lrcfetch/lrc.py
+++ b/lrcfetch/lrc.py
@@ -18,10 +18,55 @@ LRC_LINE_RE = re.compile(r"^\[(\d{2}:\d{2}[.:]\d{2,3})\]", re.MULTILINE)
 # All-zero tags
 _ZERO_TAG_RE = re.compile(r"^\[00:00[.:]0{2,3}\]$")
 
+# [offset:+/-xxx] tag — value in milliseconds
+_OFFSET_RE = re.compile(r"^\[offset:\s*([+-]?\d+)\]\s*$", re.MULTILINE | re.IGNORECASE)
+
+# Time tag for offset application: captures mm, ss, cc/ccc
+_TIME_TAG_RE = re.compile(r"\[(\d{2}):(\d{2})\.(\d{2,3})\]")
+
+
+def _apply_offset(text: str) -> str:
+    """Parse [offset:±ms] tag and shift all time tags accordingly.
+
+    Per LRC spec, a positive offset means lyrics appear sooner (subtract
+    from timestamps), negative means later (add to timestamps).
+    """
+    m = _OFFSET_RE.search(text)
+    if not m:
+        return text
+    offset_ms = int(m.group(1))
+    if offset_ms == 0:
+        return _OFFSET_RE.sub("", text).strip("\n")
+
+    # Remove the offset tag line
+    text = _OFFSET_RE.sub("", text)
+
+    def _shift(match: re.Match) -> str:
+        mm, ss, cs = int(match.group(1)), int(match.group(2)), match.group(3)
+        # Normalize centiseconds to milliseconds
+        if len(cs) == 2:
+            ms = int(cs) * 10
+            fmt_cs = 2
+        else:
+            ms = int(cs)
+            fmt_cs = 3
+        total_ms = (mm * 60 + ss) * 1000 + ms - offset_ms
+        total_ms = max(0, total_ms)
+        new_mm = total_ms // 60000
+        new_ss = (total_ms % 60000) // 1000
+        new_cs = total_ms % 1000
+        if fmt_cs == 2:
+            new_cs = new_cs // 10
+            return f"[{new_mm:02d}:{new_ss:02d}.{new_cs:02d}]"
+        return f"[{new_mm:02d}:{new_ss:02d}.{new_cs:03d}]"
+
+    return _TIME_TAG_RE.sub(_shift, text)
+
 
 def normalize_tags(text: str) -> str:
-    """Convert non-standard time tags [mm:ss:cc] to standard [mm:ss.cc]."""
-    return _COLON_TAG_RE.sub(r"[\1.\2]", text)
+    """Normalize LRC time tags: colon format → dot format, then apply offset."""
+    text = _COLON_TAG_RE.sub(r"[\1.\2]", text)
+    return _apply_offset(text)
 
 
 def is_synced(text: str) -> bool: