changes for gwen tts

.vscode/launch.json

@@ -12,4 +12,4 @@
             "module": "anki_hsk_creator"
         }
     ]
-}
+}

src/anki_hsk_creator/__main__.py

@@ -8,11 +8,12 @@ from .api import (
     is_file,
     list_input_files,
     pre_process_a_dictionary_file,
+    process_a_dictation_file,
     process_a_dictionary_file,
     process_a_phrases_file,
     select_file,
 )
-from .constants import DICT_TYPE, LANGUAGES, PHRASES_TYPE
+from .constants import DICTATION_TYPE, DICT_TYPE, LANGUAGES, PHRASES_TYPE
 from .utility import ProcessFile
 
 
@@ -90,6 +91,12 @@ def main():
                 f"processing file {input_file.input_file} with language {language_id}"
             )
             process_a_phrases_file(input_file, language_id)
+        elif DICTATION_TYPE in input_file.input_file.suffixes:
+            language_id = cli_select_language()
+            print(
+                f"processing file {input_file.input_file} with language {language_id}"
+            )
+            process_a_dictation_file(input_file, language_id)
 
 
 if __name__ == "__main__":

src/anki_hsk_creator/api.py

@@ -221,7 +221,7 @@ def process_a_dictation_file(process_file: ProcessFile, language_id: str) -> Pat
     with process_file.absolute_input_file.open(
         "r", encoding="utf8", newline="\n"
     ) as file:
-        text_lines = [line.strip() for line in file.read().split("。")]
+        text_lines = [line.strip() for line in file.read().split("。") if line.strip()]
     results = dictation_process(text_lines, process_file)
     return output_anki_dictation(process_file, results)
 
@@ -234,6 +234,6 @@ def process_a_phrases_file(process_file: ProcessFile, language_id: str) -> Path:
     with process_file.absolute_input_file.open(
         "r", encoding="utf8", newline="\n"
     ) as file:
-        text_lines = [line.strip() for line in file.readlines()]
+        text_lines = [line.strip() for line in file.readlines() if line.strip()]
     results = translator_process(text_lines, process_file)
     return output_anki_phrase(process_file, results)

src/anki_hsk_creator/constants.py

@@ -16,7 +16,7 @@ OUTPUT = DATA_FOLDER / "output"
 OUTPUT.mkdir(exist_ok=True, parents=True)
 RESOURCES = DATA_FOLDER / "resources"
 RESOURCES.mkdir(exist_ok=True, parents=True)
-GWEN_TTS = GWEN_FOLDER / "Qwen3-TTS-12Hz-1.7B-CustomVoice"
+GWEN_TTS = GWEN_FOLDER / "Qwen3-TTS-12Hz-0.6B-CustomVoice"
 
 # File Types
 PHRASES_TYPE = ".phrases"

src/anki_hsk_creator/proccessor.py

@@ -25,10 +25,10 @@ def dictation_process(
     """Process for Dictation translation"""
     results = []
     for n, line in enumerate(text_lines):
-        line = line.strip()
+        audio_line = line.strip()
-        line = "  ".join(line.split())
+        # line = "  ".join(line.split())
-        audio_line = " ".join(line)
+        # audio_line = " ".join(line)
-        audio_line = audio_line.replace("，", "，。。。]")
+        # audio_line = audio_line.replace("，", "，。。。]")
         audio_path = process_file.resources / f"N{n:03n}.wav"
         if not audio_path.exists():
             wavs, sr = TTS.generate(f"{audio_line}。")
@@ -48,10 +48,10 @@ def translator_process(
     """Process for phases or sentence translation"""
     results = []
     for n, line in enumerate(text_lines):
-        line = line.strip()
+        audio_line = line.strip()
-        line = "  ".join(line.split())
+        # line = "  ".join(line.split())
-        audio_line = " ".join(line)
+        # audio_line = " ".join(line)
-        audio_line = audio_line.replace("，", "，。。。]")
+        # audio_line = audio_line.replace("，", "，。。。]")
         audio_path = process_file.resources / f"N{n:03n}.wav"
         if not audio_path.exists():
             wavs, sr = TTS.generate(f"{audio_line}。")

src/anki_hsk_creator/utility.py

@@ -151,7 +151,7 @@ class TTS:
         if TTS.DEVICE is None:
             # Automatically detect the best available device
             if torch.cuda.is_available():
-                TTS.DEVICE = "cuda"
+                TTS.DEVICE = "cuda:0"
             elif torch.backends.mps.is_available():
                 TTS.DEVICE = "mps"
             else:
@@ -161,7 +161,7 @@ class TTS:
                 GWEN_TTS,
                 device_map=TTS.DEVICE,
                 dtype=torch.bfloat16,
-                # attn_implementation="flash_attention_2",
+                attn_implementation="flash_attention_2",
             )
 
     @staticmethod