diff --git a/picard/metadata.py b/picard/metadata.py
index dca8ffb8a..180da615a 100644
--- a/picard/metadata.py
+++ b/picard/metadata.py
@@ -433,8 +433,9 @@ class Metadata(MutableMapping):
         name = self.normalize_tag(name)
         if isinstance(values, str) or not isinstance(values, Iterable):
             values = [values]
-        values = [str(value) for value in values if value or value == 0]
-        if values:
+        values = [str(value) for value in values if value or value == 0 or value == '']
+        # Remove if there is only a single empty or blank element.
+        if values and (len(values) > 1 or values[0]):
             self._store[name] = values
             self.deleted_tags.discard(name)
         elif name in self._store:
diff --git a/picard/script/functions.py b/picard/script/functions.py
index 091a1a023..93c3a4d10 100644
--- a/picard/script/functions.py
+++ b/picard/script/functions.py
@@ -1494,3 +1494,24 @@ _Since Picard 2.7_"""
 def func_is_multi(parser, multi):
     multi_value = MultiValue(parser, multi, MULTI_VALUED_JOINER)
     return '' if len(multi_value) < 2 else '1'
+
+
+@script_function(eval_args=True, documentation=N_(
+    """`$clean_multi(name)`
+
+Removes all empty string elements from the multi-value variable.
+
+Example:
+
+    $$setmulti(test,one; ; two; three)
+    $clean_multi(%test%)
+
+Result: Sets the value of 'test' to ["one", "two", "three"].
+
+_Since Picard 2.8_"""
+))
+def func_clean_multi(parser, multi):
+    name = normalize_tagname(multi)
+    values = [str(value) for value in parser.context.getall(name) if value or value == 0]
+    parser.context[multi] = values
+    return ""
diff --git a/test/test_script.py b/test/test_script.py
index 49d18bb7e..f72714360 100644
--- a/test/test_script.py
+++ b/test/test_script.py
@@ -427,16 +427,16 @@ class ScriptParserTest(PicardTestCase):
         self.assertEqual("", self.parser.eval("$setmulti(test,%source%)", context))  # no return value
         self.assertEqual(context.getall("source"), context.getall("test"))
 
-    def test_cmd_setmulti_will_remove_empty_items(self):
+    def test_cmd_setmulti_will_keep_empty_items(self):
         context = Metadata()
         context["source"] = ["", "multi", ""]
         self.assertEqual("", self.parser.eval("$setmulti(test,%source%)", context))  # no return value
-        self.assertEqual(["multi"], context.getall("test"))
+        self.assertEqual(["", "multi", ""], context.getall("test"))
 
     def test_cmd_setmulti_custom_splitter_string(self):
         context = Metadata()
         self.assertEqual("", self.parser.eval("$setmulti(test,multi##valued##test##,##)", context))  # no return value
-        self.assertEqual(["multi", "valued", "test"], context.getall("test"))
+        self.assertEqual(["multi", "valued", "test", ""], context.getall("test"))
 
     def test_cmd_setmulti_empty_splitter_does_nothing(self):
         context = Metadata()
@@ -1785,3 +1785,56 @@ class ScriptParserTest(PicardTestCase):
             self.parser.eval("$is_multi()")
         with self.assertRaisesRegex(ScriptError, areg):
             self.parser.eval("$is_multi(a,)")
+
+    def test_cmd_clean_multi(self):
+        context = Metadata()
+        context["foo"] = ["", "one", "two"]
+        context["bar"] = ["one", "", "two"]
+        context["baz"] = ["one", "two", ""]
+
+        # Confirm initial values
+        self.assertScriptResultEquals("%foo%", "; one; two", context)
+        self.assertScriptResultEquals("%bar%", "one; ; two", context)
+        self.assertScriptResultEquals("%baz%", "one; two; ", context)
+        # Test cleaned values
+        self.assertScriptResultEquals("$clean_multi(foo)%foo%", "one; two", context)
+        self.assertScriptResultEquals("$clean_multi(bar)%bar%", "one; two", context)
+        self.assertScriptResultEquals("$clean_multi(baz)%baz%", "one; two", context)
+
+        # Text clean with only empty string elements
+        context["foo"] = ["", "", ""]
+
+        # Confirm initial values
+        self.assertScriptResultEquals("%foo%", "; ; ", context)
+        # Test cleaned values
+        self.assertScriptResultEquals("$clean_multi(foo)%foo%", "", context)
+
+        # Test clean with indirect argument
+        context["foo"] = ["", "one", "two"]
+        context["bar"] = "foo"
+
+        # Confirm initial values
+        self.assertScriptResultEquals("%foo%", "; one; two", context)
+        # Test cleaned values
+        self.assertScriptResultEquals("$clean_multi(%bar%)%foo%", "one; two", context)
+
+        # Test clean with non-multi argument
+        context["foo"] = "one"
+        context["bar"] = "one; ; two"
+        context["baz"] = ""
+
+        # Confirm initial values
+        self.assertScriptResultEquals("%foo%", "one", context)
+        self.assertScriptResultEquals("%bar%", "one; ; two", context)
+        self.assertScriptResultEquals("%baz%", "", context)
+        # Test cleaned values
+        self.assertScriptResultEquals("$clean_multi(foo)%foo%", "one", context)
+        self.assertScriptResultEquals("$clean_multi(bar)%bar%", "one; ; two", context)
+        self.assertScriptResultEquals("$clean_multi(baz)%baz%", "", context)
+
+        # Tests with invalid number of arguments
+        areg = r"^\d+:\d+:\$clean_multi: Wrong number of arguments for \$clean_multi: Expected exactly 1, "
+        with self.assertRaisesRegex(ScriptError, areg):
+            self.parser.eval("$clean_multi()")
+        with self.assertRaisesRegex(ScriptError, areg):
+            self.parser.eval("$clean_multi(foo,)")