diff --git a/CHANGELOG.rst b/CHANGELOG.rst
index db17ef752..4a481de97 100644
--- a/CHANGELOG.rst
+++ b/CHANGELOG.rst
@@ -44,6 +44,7 @@ Features:
 Fixes:
 
 - Support slicing ``Container.streams``, which raised ``TypeError`` despite being typed.
+- ``AssSubtitle.dialogue`` now strips override tags such as ``{\b1}``, turns ``\n`` and ``\h`` into spaces, and keeps a ``{`` with no closing ``}`` instead of dropping the rest of the line.
 - ``HWConfig.methods`` no longer raises ``ValueError`` for configs that support several methods; ``HWConfigMethod`` is now an ``IntFlag``.
 - ``CodecContext.is_encoder``, ``CodecContext.is_decoder``, ``Container.input_was_opened``, and ``VideoFormatComponent.is_chroma`` now return ``bool`` instead of ``int``.
 - ``Frame.time_base`` returns ``AVRational(0, 1)`` instead of ``None`` when unset, like other ``AVRational`` properties.
diff --git a/av/subtitles/subtitle.py b/av/subtitles/subtitle.py
index 2e45b4e3a..bd2629fa4 100644
--- a/av/subtitles/subtitle.py
+++ b/av/subtitles/subtitle.py
@@ -291,15 +291,21 @@ def ass(self):
     @property
     def dialogue(self):
         """
-        Extract the dialogue from the ass format. Strip comments.
+        Extract the plain text from the ASS dialogue line.
+
+        Override tags and comments (``{...}`` blocks) are removed, ``\\N``
+        becomes a newline, and ``\\n`` and ``\\h`` become a space. A ``{``
+        with no closing ``}`` is kept as literal text.
         """
+        ass_text: bytes = self.ass
+        text_len: cython.Py_ssize_t = len(ass_text)
+        # A '{' only opens a block if a '}' appears somewhere after it.
+        last_close: cython.Py_ssize_t = ass_text.rfind(b"}")
         comma_count: cython.short = 0
         i: cython.Py_ssize_t = 0
-        state: cython.bint = False
-        ass_text: bytes = self.ass
+        in_block: cython.bint = False
         char, next_char = cython.declare(cython.uchar)
         result: bytearray = bytearray()
-        text_len: cython.Py_ssize_t = len(ass_text)
 
         while comma_count < 8 and i < text_len:
             if ass_text[i] == b","[0]:
@@ -310,18 +316,21 @@ def dialogue(self):
             char = ass_text[i]
             next_char = 0 if i + 1 >= text_len else ass_text[i + 1]
 
-            if char == b"\\"[0] and next_char == b"N"[0]:
-                result.append(b"\n"[0])
-                i += 2
-                continue
-
-            if not state:
-                if char == b"{"[0] and next_char != b"\\"[0]:
-                    state = True
+            if not in_block:
+                if char == b"\\"[0] and next_char == b"N"[0]:
+                    result.append(b"\n"[0])
+                    i += 2
+                    continue
+                if char == b"\\"[0] and (next_char == b"n"[0] or next_char == b"h"[0]):
+                    result.append(b" "[0])
+                    i += 2
+                    continue
+                if char == b"{"[0] and i < last_close:
+                    in_block = True
                 else:
                     result.append(char)
             elif char == b"}"[0]:
-                state = False
+                in_block = False
             i += 1
 
         return bytes(result)
diff --git a/tests/test_subtitles.py b/tests/test_subtitles.py
index b5b70746a..51cf3f3b2 100644
--- a/tests/test_subtitles.py
+++ b/tests/test_subtitles.py
@@ -154,6 +154,28 @@ def test_subtitle_dialogue_extended_chars(self) -> None:
 
         assert sub.dialogue == "♪ Hey, hey, hey ♪".encode()
 
+    def test_subtitle_dialogue(self) -> None:
+        from av.subtitles.subtitle import SubtitleSet
+
+        def dialogue(text: bytes) -> bytes:
+            subtitle = SubtitleSet.create(text=text, start=0, end=2000)
+            return cast(AssSubtitle, subtitle[0]).dialogue
+
+        prefix = b"0,0,Default,,0,0,0,,"
+        # Override tags and comments are stripped.
+        assert dialogue(prefix + b"{\\b1}bold{\\b0}") == b"bold"
+        assert dialogue(prefix + b"{\\an8}top") == b"top"
+        assert dialogue(prefix + b"{\\fnArial}hi") == b"hi"
+        assert dialogue(prefix + b"a{comment}b") == b"ab"
+        # Text escapes: \N -> newline, \n and \h -> space.
+        assert dialogue(prefix + b"a\\Nb") == b"a\nb"
+        assert dialogue(prefix + b"a\\nb") == b"a b"
+        assert dialogue(prefix + b"a\\hb") == b"a b"
+        # A stray '{' with no closing '}' is literal text.
+        assert dialogue(prefix + b"cost is {5") == b"cost is {5"
+        # Commas in the Text field are kept.
+        assert dialogue(prefix + b"x, y, z") == b"x, y, z"
+
     def test_subtitle_encode_mp4(self) -> None:
         """Test encoding subtitles to MP4 container."""
         from av.subtitles.subtitle import SubtitleSet