Normalize punctuation before German TTS

platform/docker/tts-gateway/test_tts_gateway.py

@@ -110,6 +110,21 @@ class LanguageSegmentationTests(unittest.TestCase):
                 self.assertIn("29 Kilometer pro Stunde", spoken)
                 self.assertNotIn("km", spoken.lower())
 
+    def test_visual_punctuation_becomes_natural_pauses(self):
+        spoken = gateway.clean_for_speech(
+            "Status: stabil – keine Fehler; Docker-Container laufen."
+        )
+        self.assertEqual(
+            spoken,
+            "Status, stabil, keine Fehler, Docker Container laufen.",
+        )
+        self.assertNotRegex(spoken, r"[:;\-‐‑‒–—−]")
+
+    def test_punctuation_only_segments_are_never_synthesized(self):
+        segments = gateway.prepare_segments("Status: – alles läuft.")
+        self.assertTrue(segments)
+        self.assertTrue(all(any(char.isalnum() for char in part) for _, part in segments))
+
     def test_weather_summary_is_split_into_short_complete_chunks(self):
         text = (
             "Heute in Rastatt: teils sonnig, trocken, 10 bis 22 Grad. "

platform/docker/tts-gateway/tts_gateway.py

@@ -290,16 +290,22 @@ def clean_for_speech(text: str) -> str:
     text = re.sub(r"(?m)^\s{0,3}#{1,6}\s*", "", text)
     text = re.sub(r"(?m)^\s*[-*+]\s+", "", text)
     text = text.replace("→", ". ").replace("←", ". ")
-    text = text.replace("–", " - ").replace("—", " - ")
     if DEFAULT_LANGUAGE == "de":
         text = normalize_for_german_speech(text)
+    # XTTS occasionally hallucinates syllables when visual punctuation is
+    # submitted literally or isolated at a chunk boundary. Preserve its pause
+    # semantics, but never ask the model to pronounce the glyph itself.
+    text = re.sub(r"(?<=\w)[\-‐‑‒–—−](?=\w)", " ", text)
+    text = re.sub(r"\s+[\-‐‑‒–—−]\s+", ", ", text)
+    text = re.sub(r"\s*[:;]+\s*", ", ", text)
     text = "".join(
         char for char in text
         if unicodedata.category(char) not in {"So", "Cs"}
     )
     text = re.sub(r"[ \t]+", " ", text)
     text = re.sub(r"\s*\n+\s*", ". ", text)
-    text = re.sub(r"([:;])\s*\.", r"\1", text)
+    text = re.sub(r",\s*\.", ".", text)
+    text = re.sub(r"(?:,\s*){2,}", ", ", text)
     text = re.sub(r"(?:\.\s*){2,}", ". ", text)
     return text.strip()