add audio generation suport

anki-hsk-creator/__init__.py

@@ -1,2 +1,5 @@
 """anki-hsk-creator"""
-HF_TOKEN = "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"
+
+import os
+
+os.environ["HF_TOKEN"] = "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"

anki-hsk-creator/__main__.py

@@ -10,6 +10,8 @@ from genanki import Deck, Note, Model, Package
 import argostranslate.package
 import argostranslate.translate
 from chatterbox.mtl_tts import ChatterboxMultilingualTTS
+import torch
+import torchaudio
 
 ## Constants
 
@@ -68,7 +70,7 @@ HSK_FRONT_TEMPLATE = """
 {{English}}
 </tts>
 <br>
-{{MyMedia}}
+{{Audio}}
 """
 
 HSK_MODEL = Model(
@@ -92,7 +94,7 @@ HSK_MODEL = Model(
             "name": "Card 2",
             "qfmt": "<div class='simple'>{{Simplified}}</div><br><div class='trad'>"
             "{{Traditional}}</div>",
-            "afmt": '{{FrontSide}}<hr id="answer"><strong>{{Pinyin}}</strong><br>{{English}}<br>{{MyMedia}}',
+            "afmt": '{{FrontSide}}<hr id="answer"><strong>{{Pinyin}}</strong><br>{{English}}<br>{{Audio}}',
         },
     ],
     css=CSS,
@@ -128,7 +130,14 @@ def create_translator():
     argostranslate.package.install_from_path(package_to_install.download())
 
 def create_tts():
-    tts = ChatterboxMultilingualTTS.from_pretrained(device="cuda")
+    # Automatically detect the best available device
+    if torch.cuda.is_available():
+        device = "cuda"
+    elif torch.backends.mps.is_available():
+        device = "mps"
+    else:
+        device = "cpu"
+    tts = ChatterboxMultilingualTTS.from_pretrained(device=device)
     return tts
 
 ## Main
@@ -166,6 +175,7 @@ def dictionary_process(dictionary, tts, in_file, resources):
     """Process dictionary files"""
     words_list = in_file.open(encoding="utf8").read().strip().split("\n")
     results = [] 
+    try:
         with in_file.open("w", encoding="utf8") as input_file:
             for words in words_list:
                 word = words.split()[0]
@@ -190,9 +200,10 @@ def dictionary_process(dictionary, tts, in_file, resources):
                             v = ml[0]
                     else:
                         v = v[0]
-                audio = tts.generate(word, language_id="zh")
                     audio_path = resources / f"{word}.wav"
-                ta.save(audio_path, audio, tts.sr)
+                    if not audio_path.exists():
+                        audio = tts.generate(word, language_id="zh")
+                        torchaudio.save(audio_path, audio, tts.sr)
                     input_file.write(f"{word}\t{v.pinyin}\n")
                     results.append((v, audio_path))
                 else:
@@ -200,6 +211,9 @@ def dictionary_process(dictionary, tts, in_file, resources):
                     print(f"===================>ERROR: {word} not found")
                     print("============================================")
                     input_file.write(f"{word}\tERROR\n")
+    except Exception:
+        with in_file.open("w", encoding="utf8") as input_file:
+            input_file.write("\n".join(words_list))
     return results
 
 

anki_hsk_creator.egg-info/PKG-INFO

@@ -7,5 +7,8 @@ Requires-Dist: pinyin-tone-converter
 Requires-Dist: genanki
 Requires-Dist: argostranslate
 Requires-Dist: chatterbox-tts
+Requires-Dist: torch
+Requires-Dist: torchaudio
+Requires-Dist: torchcodec
 Dynamic: license-file
 Dynamic: requires-dist

anki_hsk_creator.egg-info/requires.txt

@@ -3,3 +3,6 @@ pinyin-tone-converter
 genanki
 argostranslate
 chatterbox-tts
+torch
+torchaudio
+torchcodec

setup.py

@@ -10,6 +10,8 @@ setup(
         "genanki",
         "argostranslate",
         "chatterbox-tts",
+        "torch",
         "torchaudio",
+        "torchcodec"
     ],
 )