refactor(webui): use shared get_split_pattern instead of hardcoded \\n+

All three Web UI consumers now call domain.split_pattern.get_split_pattern()
which selects the correct split pattern based on language and subtitle mode.

Before: WebUI always split on \\n+ regardless of language (CJK missed
punctuation-based splitting that PyQt already had).
After: Both UIs share identical language-aware splitting logic.

1038 tests pass.

abogen/webui/conversion_runner.py

@@ -111,6 +111,7 @@ from abogen.domain.output_paths import (
     resolve_project_layout as _resolve_project_layout,
 )
 from abogen.domain.device import select_device as _select_device
+from abogen.domain.split_pattern import get_split_pattern
 from abogen.domain.audio_helpers import (
     build_ffmpeg_command as _build_ffmpeg_command,
     to_float32 as _to_float32,
@@ -127,7 +128,7 @@ from .service import Job, JobStatus
 
 _export_svc = ExportService()
 
-SPLIT_PATTERN = r"\n+"
+SPLIT_PATTERN = r"\n+"  # Kept for backward compatibility; prefer get_split_pattern()
 SAMPLE_RATE = 24000
 
 
@@ -172,6 +173,12 @@ def run_conversion_job(job: Job) -> None:
                 "LLM-based apostrophe normalization is selected, but the LLM configuration is incomplete."
             )
 
+    # Compute language-aware split pattern once for the entire job
+    job_split_pattern = get_split_pattern(
+        str(job.language or "a"),
+        str(job.subtitle_mode or "Disabled"),
+    )
+
     sink_stack = ExitStack()
     subtitle_writer = None
     chapter_paths: list[Path] = []
@@ -438,12 +445,14 @@ def run_conversion_job(job: Job) -> None:
             voice_choice: Any,
             chapter_sink: Optional[AudioSink],
             preview_prefix: Optional[str] = None,
-            split_pattern: Optional[str] = SPLIT_PATTERN,
+            split_pattern: Optional[str] = None,
             tts_provider: Optional[str] = None,
             speed_override: Optional[float] = None,
             supertonic_steps_override: Optional[int] = None,
         ) -> int:
             nonlocal processed_chars, current_time
+            if split_pattern is None:
+                split_pattern = job_split_pattern
             source_text = str(text or "")
             try:
                 normalized = prepare_text_for_tts(
@@ -625,7 +634,6 @@ def run_conversion_job(job: Job) -> None:
                         voice_choice=voice_choice,
                         chapter_sink=chapter_sink,
                         preview_prefix=f"Chapter {idx} title",
-                        split_pattern=SPLIT_PATTERN,
                         tts_provider=chapter_provider,
                         speed_override=chapter_speed,
                         supertonic_steps_override=chapter_steps,

abogen/webui/debug_tts_runner.py

@@ -14,7 +14,8 @@ from abogen.kokoro_text_normalization import normalize_for_pipeline
 from abogen.normalization_settings import build_apostrophe_config
 from abogen.text_extractor import extract_from_path
 from abogen.voice_cache import ensure_voice_assets
-from abogen.webui.conversion_runner import SAMPLE_RATE, SPLIT_PATTERN, _select_device, _to_float32, _resolve_voice, _spec_to_voice_ids
+from abogen.webui.conversion_runner import SAMPLE_RATE, _select_device, _to_float32, _resolve_voice, _spec_to_voice_ids
+from abogen.domain.split_pattern import get_split_pattern
 from abogen.tts_plugin.utils import create_pipeline
 
 
@@ -200,7 +201,7 @@ def run_debug_tts_wavs(
             normalized,
             voice=voice_choice,
             speed=speed,
-            split_pattern=SPLIT_PATTERN,
+            split_pattern=get_split_pattern(language, "Disabled"),
         ):
             audio = _to_float32(getattr(segment, "audio", None))
             if audio.size:

abogen/webui/routes/utils/synthesize.py

@@ -7,9 +7,9 @@ from flask import current_app, send_file
 from flask.typing import ResponseReturnValue
 
 from abogen.domain.device import select_device as _select_device
+from abogen.domain.split_pattern import get_split_pattern
 
 
-SPLIT_PATTERN = r"\n+"
 SAMPLE_RATE = 24000
 
 _preview_pipelines: Dict[Tuple[str, str], Any] = {}
@@ -123,6 +123,8 @@ def generate_preview_audio(
             current_app.logger.exception("Preview normalization failed; using raw text")
             normalized_text = source_text
 
+    preview_split = get_split_pattern(str(language or "a"), "Disabled")
+
     if provider == "supertonic":
         from abogen.tts_plugin.utils import create_pipeline
 
@@ -131,7 +133,7 @@ def generate_preview_audio(
             normalized_text,
             voice=voice_spec,
             speed=speed,
-            split_pattern=SPLIT_PATTERN,
+            split_pattern=preview_split,
             total_steps=supertonic_total_steps,
         )
     else:
@@ -149,7 +151,7 @@ def generate_preview_audio(
             normalized_text,
             voice=voice_choice,
             speed=speed,
-            split_pattern=SPLIT_PATTERN,
+            split_pattern=preview_split,
         )
 
     audio_chunks: List[np.ndarray] = []