From e77c8b3372d48e7f3f3d94e49a49ae39bc1f8c09 Mon Sep 17 00:00:00 2001
From: Artem Akymenko <akimenko.artem@gmail.com>
Date: Sat, 18 Jul 2026 06:52:28 +0000
Subject: [PATCH] =?UTF-8?q?fix:=20review=20cleanup=20=E2=80=94=20imports,?=
 =?UTF-8?q?=20=5FFakeToken,=20use=5Fspacy=5Fsegmentation?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- Move pronunciation imports from inside run() to top-level imports
- Extract _FakeToken to module level (was redefined every loop iteration)
- use_spacy_segmentation now mirrors PyQt logic: pass the flag,
  let process_subtitle_tokens filter by language internally
---
 abogen/pyqt/conversion.py         | 10 +++++-----
 abogen/webui/conversion_runner.py | 19 ++++++++++++-------
 2 files changed, 17 insertions(+), 12 deletions(-)

diff --git a/abogen/pyqt/conversion.py b/abogen/pyqt/conversion.py
index 75604b0..3a139c2 100644
--- a/abogen/pyqt/conversion.py
+++ b/abogen/pyqt/conversion.py
@@ -39,6 +39,11 @@ from abogen.domain.subtitle_generation import process_subtitle_tokens
 from abogen.domain.voice_loader import load_voice_cached
 from abogen.domain.progress import calc_etr_str
 from abogen.domain.normalization import prepare_text_for_tts
+from abogen.domain.pronunciation import (
+    compile_pronunciation_rules,
+    compile_heteronym_sentence_rules,
+    merge_pronunciation_overrides,
+)
 from abogen.domain.metadata_extraction import (
     extract_metadata_and_build_args,
     read_text_for_metadata,
@@ -537,11 +542,6 @@ class ConversionThread(QThread):
                 )
 
             # --- Compile normalization rules (heteronym + pronunciation) ---
-            from abogen.domain.pronunciation import (
-                compile_pronunciation_rules,
-                compile_heteronym_sentence_rules,
-                merge_pronunciation_overrides,
-            )
             pronunciation_overrides = merge_pronunciation_overrides(
                 getattr(self, "pronunciation_overrides", None),
                 getattr(self, "manual_overrides", None),
diff --git a/abogen/webui/conversion_runner.py b/abogen/webui/conversion_runner.py
index 464fa2c..5fd1afc 100644
--- a/abogen/webui/conversion_runner.py
+++ b/abogen/webui/conversion_runner.py
@@ -135,6 +135,16 @@ SPLIT_PATTERN = r"\n+"  # Kept for backward compatibility; prefer get_split_patt
 SAMPLE_RATE = 24000
 
 
+class _FakeToken:
+    """Minimal token stub for languages without per-word token support."""
+
+    def __init__(self, text: str, start: float, end: float):
+        self.text = text
+        self.start_ts = start
+        self.end_ts = end
+        self.whitespace = ""
+
+
 class _JobCancelled(Exception):
     """Raised internally to abort a conversion when the client cancels."""
 
@@ -534,12 +544,6 @@ def run_conversion_job(job: Job) -> None:
 
                         # Fallback for languages without token support: create a single token
                         if not tokens_list and graphemes:
-                            class _FakeToken:
-                                def __init__(self, text, start, end):
-                                    self.text = text
-                                    self.start_ts = start
-                                    self.end_ts = end
-                                    self.whitespace = ""
                             tokens_list = [_FakeToken(graphemes, 0, duration)]
 
                         for tok in tokens_list:
@@ -555,6 +559,7 @@ def run_conversion_job(job: Job) -> None:
 
                 # Flush accumulated tokens through process_subtitle_tokens
                 if subtitle_writer and audio_sink and accumulated_tokens:
+                    _use_spacy = job.subtitle_mode not in ("Disabled", "Line")
                     new_entries: List[tuple] = []
                     process_subtitle_tokens(
                         accumulated_tokens,
@@ -562,7 +567,7 @@ def run_conversion_job(job: Job) -> None:
                         job.max_subtitle_words,
                         job.subtitle_mode,
                         job.language,
-                        use_spacy_segmentation=False,
+                        use_spacy_segmentation=_use_spacy,
                         fallback_end_time=current_time,
                     )
                     for start, end, text in new_entries: