Normalize structured German TTS input

platform/docker/tts-gateway/tts_gateway.py

@@ -116,6 +116,16 @@ ENGLISH_PRONUNCIATIONS = {
     "gib": "gigabytes",
 }
 
+GERMAN_MONTHS = {
+    1: "Januar", 2: "Februar", 3: "März", 4: "April",
+    5: "Mai", 6: "Juni", 7: "Juli", 8: "August",
+    9: "September", 10: "Oktober", 11: "November", 12: "Dezember",
+}
+GERMAN_DIGITS = {
+    "0": "null", "1": "eins", "2": "zwei", "3": "drei", "4": "vier",
+    "5": "fünf", "6": "sechs", "7": "sieben", "8": "acht", "9": "neun",
+}
+
 
 def _request(url: str, *, payload: dict | None = None,
              timeout: float = 10) -> tuple[bytes, str]:
@@ -190,6 +200,74 @@ def segment_languages(text: str) -> list[tuple[str, str]]:
     return merged
 
 
+def _spoken_date(match: re.Match) -> str:
+    day = int(match.group(1))
+    month = int(match.group(2))
+    year = match.group(3)
+    month_name = GERMAN_MONTHS.get(month)
+    if not month_name or not 1 <= day <= 31:
+        return match.group(0)
+    result = f"{day}. {month_name}"
+    if year:
+        result += f" {year}"
+    return result
+
+
+def _spell_digits(value: str) -> str:
+    return " ".join(GERMAN_DIGITS[digit] for digit in value)
+
+
+def normalize_for_german_speech(text: str) -> str:
+    """Turn common visual notation into unambiguous spoken German."""
+    text = re.sub(
+        r"(-?\d+(?:[,.]\d+)?)[ \t]*°[ \t]*(?:C)?[ \t]*/[ \t]*"
+        r"(-?\d+(?:[,.]\d+)?)[ \t]*°[ \t]*(?:C)?",
+        r"Höchstwert \1 Grad, Tiefstwert \2 Grad",
+        text,
+        flags=re.IGNORECASE,
+    )
+    text = re.sub(
+        r"(-?\d+(?:[,.]\d+)?)[ \t]*°[ \t]*(?:C)?",
+        r"\1 Grad",
+        text,
+        flags=re.IGNORECASE,
+    )
+    text = re.sub(
+        r"\b([0-3]?\d)\.([01]?\d)\.(?:(\d{4})\b)?",
+        _spoken_date,
+        text,
+    )
+    text = re.sub(
+        r"\bPLZ\s+(\d{5})\b",
+        lambda match: "Postleitzahl " + _spell_digits(match.group(1)),
+        text,
+        flags=re.IGNORECASE,
+    )
+    text = re.sub(
+        r"\((\d{5})\)",
+        lambda match: "(Postleitzahl " + _spell_digits(match.group(1)) + ")",
+        text,
+    )
+    text = re.sub(
+        r"https?://(?:www\.)?([^/\s)]+)(?:/[^\s)]*)?",
+        lambda match: match.group(1),
+        text,
+        flags=re.IGNORECASE,
+    )
+    text = re.sub(
+        r"\b([A-Za-z0-9][A-Za-z0-9-]*(?:\.[A-Za-z0-9-]+)*)"
+        r"\.(de|com|org|net|io|ai)\b",
+        lambda match: (
+            match.group(1).replace(".", " Punkt ")
+            + " Punkt " + " ".join(match.group(2).upper())
+        ),
+        text,
+        flags=re.IGNORECASE,
+    )
+    text = re.sub(r"(\d)\s*%", r"\1 Prozent", text)
+    return text
+
+
 def clean_for_speech(text: str) -> str:
     """Remove visual markup that makes long TTS output unstable or noisy."""
     text = re.sub(r"```.*?```", " Code block. ", text, flags=re.DOTALL)
@@ -200,12 +278,15 @@ def clean_for_speech(text: str) -> str:
     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)
     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*){2,}", ". ", text)
     return text.strip()