diff --git a/tests/tokenize_rt_test.py b/tests/tokenize_rt_test.py
index b6f572a..dccaa3a 100644
--- a/tests/tokenize_rt_test.py
+++ b/tests/tokenize_rt_test.py
@@ -201,6 +201,26 @@ def test_src_to_tokens_fstring_with_escapes():
         ]
 
 
+def test_src_to_tokens_fstring_with_named_escapes():
+    src = r'f" \N{SNOWMAN} "'
+    ret = src_to_tokens(src)
+    if sys.version_info >= (3, 12):  # pragma: >=3.12 cover
+        assert ret == [
+            Token(name='FSTRING_START', src='f"', line=1, utf8_byte_offset=0),
+            Token(name='FSTRING_MIDDLE', src=' \\N{SNOWMAN}', line=1, utf8_byte_offset=2),  # noqa: E501
+            Token(name='FSTRING_MIDDLE', src=' ', line=1, utf8_byte_offset=14),
+            Token(name='FSTRING_END', src='"', line=1, utf8_byte_offset=15),
+            Token(name='NEWLINE', src='', line=1, utf8_byte_offset=16),
+            Token(name='ENDMARKER', src='', line=2, utf8_byte_offset=0),
+        ]
+    else:  # pragma: <3.12 cover
+        assert ret == [
+            Token(name='STRING', src='f" \\N{SNOWMAN} "', line=1, utf8_byte_offset=0),  # noqa: E501
+            Token(name='NEWLINE', src='', line=1, utf8_byte_offset=16),
+            Token(name='ENDMARKER', src='', line=2, utf8_byte_offset=0),
+        ]
+
+
 @pytest.mark.parametrize(
     'filename',
     (
diff --git a/tokenize_rt.py b/tokenize_rt.py
index 559b6d5..f909ead 100644
--- a/tokenize_rt.py
+++ b/tokenize_rt.py
@@ -47,6 +47,16 @@ def matches(self, *, name: str, src: str) -> bool:
 _string_re = re.compile('^([^\'"]*)(.*)$', re.DOTALL)
 _escaped_nl_re = re.compile(r'\\(\n|\r\n|\r)')
 
+NAMED_UNICODE_RE = re.compile(r'(?<!\\)(?:\\\\)*(\\N\{[^}]+\})')
+
+
+def curly_escape(s: str) -> str:
+    parts = NAMED_UNICODE_RE.split(s)
+    return ''.join(
+        part.replace('{', '{{').replace('}', '}}') if i % 2 == 0 else part
+        for i, part in enumerate(parts)
+    )
+
 
 def _re_partition(regex: Pattern[str], s: str) -> tuple[str, str, str]:
     match = regex.search(s)
@@ -101,8 +111,10 @@ def src_to_tokens(src: str) -> list[Token]:
         tok_name = tokenize.tok_name[tok_type]
 
         if tok_name == 'FSTRING_MIDDLE':  # pragma: >=3.12 cover
-            ecol += tok_text.count('{') + tok_text.count('}')
-            tok_text = tok_text.replace('{', '{{').replace('}', '}}')
+            if '{' in tok_text or '}' in tok_text:
+                new_tok_text = curly_escape(tok_text)
+                ecol += len(new_tok_text) - len(tok_text)
+                tok_text = new_tok_text
 
         tokens.append(Token(tok_name, tok_text, sline, end_offset))
         last_line, last_col = eline, ecol