Normalize punctuation before German TTS

2026-08-24 19:26:07 +02:00
parent faba681153
commit dec4709ae9
2 changed files with 23 additions and 2 deletions
@@ -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()