Fix: merge zh_recogn & sensevoice to FunAsr

test.json

@@ -0,0 +1 @@
+{"metadata": {"transaction_key": "deprecated", "request_id": "e69753ae-4e80-4077-9ccf-54d1e32a8aa7", "sha256": "696ddf1d22a425d3a9466b2368cdd24364c9cf9de83bae0883c8d433a5389fb6", "created": "2024-11-01T06:30:17.667Z", "duration": 11.088, "channels": 1, "models": ["2a000f0d-afa1-4f20-8481-94f8bc30b0b2"], "model_info": {"2a000f0d-afa1-4f20-8481-94f8bc30b0b2": {"name": "large-whisper", "version": "2022-12-08.0", "arch": "whisper"}}}, "results": {"channels": [{"alternatives": [{"transcript": "\u53e4\u8001\u65b0\u6232\u4e2d\u767c\u73fe\u4e86\u6709\u6a5f\u5206\u5b50\u6211\u5011\u96e2\u7b2c\u4e09\u985e\u63a5\u89f8\u9084\u6709\u591a\u9060\u5fae\u535a\u6b63\u5f0f\u5c55\u958b\u62cd\u651d\u4efb\u52d9\u5df2\u7d93\u5c46\u6eff\u9031\u5e74\u6700\u8fd1\u4e5f\u50b3\u56de\u4f86\u8a31\u591a\u904e\u53bb\u96e3\u4ee5\u62cd\u651d\u5230\u7684\u7167\u7247", "confidence": 0.9891718, "words": [{"word": "\u53e4", "start": 0.06, "end": 0.22, "confidence": 0.24031715, "punctuated_word": "\u53e4"}, {"word": "\u8001", "start": 0.22, "end": 0.45999998, "confidence": 0.4851624, "punctuated_word": "\u8001"}, {"word": "\u65b0", "start": 0.45999998, "end": 0.52, "confidence": 0.9871694, "punctuated_word": "\u65b0"}, {"word": "\u6232", "start": 0.52, "end": 0.71999997, "confidence": 0.689627, "punctuated_word": "\u6232"}, {"word": "\u4e2d", "start": 0.71999997, "end": 0.94, "confidence": 0.7163843, "punctuated_word": "\u4e2d"}, {"word": "\u767c\u73fe", "start": 0.94, "end": 1.4, "confidence": 0.99306655, "punctuated_word": "\u767c\u73fe"}, {"word": "\u4e86", "start": 1.4, "end": 1.5, "confidence": 0.99730957, "punctuated_word": "\u4e86"}, {"word": "\u6709", "start": 1.5, "end": 1.7199999, "confidence": 0.9590886, "punctuated_word": "\u6709"}, {"word": "\u6a5f", "start": 1.7199999, "end": 1.88, "confidence": 0.99941957, "punctuated_word": "\u6a5f"}, {"word": "\u5206", "start": 1.88, "end": 2.08, "confidence": 0.055190008, "punctuated_word": "\u5206"}, {"word": "\u5b50", "start": 2.08, "end": 2.58, "confidence": 0.96733665, "punctuated_word": "\u5b50"}, {"word": "\u6211\u5011", "start": 2.6599998, "end": 2.8999999, "confidence": 0.9208613, "punctuated_word": "\u6211\u5011"}, {"word": "\u96e2", "start": 2.8999999, "end": 3.12, "confidence": 0.9793457, "punctuated_word": "\u96e2"}, {"word": "\u7b2c\u4e09", "start": 3.12, "end": 3.3799999, "confidence": 0.9081615, "punctuated_word": "\u7b2c\u4e09"}, {"word": "\u985e", "start": 3.3799999, "end": 3.6, "confidence": 0.99989676, "punctuated_word": "\u985e"}, {"word": "\u63a5", "start": 3.6, "end": 3.76, "confidence": 0.9999919, "punctuated_word": "\u63a5"}, {"word": "\u89f8", "start": 3.76, "end": 4.04, "confidence": 0.7521827, "punctuated_word": "\u89f8"}, {"word": "\u9084\u6709", "start": 4.04, "end": 4.3199997, "confidence": 0.9909378, "punctuated_word": "\u9084\u6709"}, {"word": "\u591a", "start": 4.3199997, "end": 4.48, "confidence": 0.17423469, "punctuated_word": "\u591a"}, {"word": "\u9060", "start": 4.48, "end": 4.98, "confidence": 0.9891718, "punctuated_word": "\u9060"}, {"word": "\u5fae", "start": 5.02, "end": 5.2, "confidence": 0.99999905, "punctuated_word": "\u5fae"}, {"word": "\u535a", "start": 5.2, "end": 5.38, "confidence": 0.9900187, "punctuated_word": "\u535a"}, {"word": "\u6b63", "start": 5.38, "end": 5.54, "confidence": 0.9524369, "punctuated_word": "\u6b63"}, {"word": "\u5f0f", "start": 5.54, "end": 5.68, "confidence": 0.9987935, "punctuated_word": "\u5f0f"}, {"word": "\u5c55", "start": 5.68, "end": 5.8599997, "confidence": 0.9839077, "punctuated_word": "\u5c55"}, {"word": "\u958b", "start": 5.8599997, "end": 6.04, "confidence": 0.9992906, "punctuated_word": "\u958b"}, {"word": "\u62cd", "start": 6.04, "end": 6.18, "confidence": 0.99999917, "punctuated_word": "\u62cd"}, {"word": "\u651d", "start": 6.18, "end": 6.3799996, "confidence": 0.99359775, "punctuated_word": "\u651d"}, {"word": "\u4efb", "start": 6.3799996, "end": 6.5, "confidence": 0.13850875, "punctuated_word": "\u4efb"}, {"word": "\u52d9", "start": 6.5, "end": 6.7599998, "confidence": 0.9938491, "punctuated_word": "\u52d9"}, {"word": "\u5df2\u7d93", "start": 6.7599998, "end": 6.96, "confidence": 0.9992866, "punctuated_word": "\u5df2\u7d93"}, {"word": "\u5c46", "start": 6.96, "end": 7.12, "confidence": 0.9860264, "punctuated_word": "\u5c46"}, {"word": "\u6eff", "start": 7.12, "end": 7.3599997, "confidence": 0.9997645, "punctuated_word": "\u6eff"}, {"word": "\u9031", "start": 7.3599997, "end": 7.54, "confidence": 0.33030534, "punctuated_word": "\u9031"}, {"word": "\u5e74", "start": 7.54, "end": 8.0, "confidence": 0.9914724, "punctuated_word": "\u5e74"}, {"word": "\u6700\u8fd1", "start": 8.0, "end": 8.3, "confidence": 0.97437614, "punctuated_word": "\u6700\u8fd1"}, {"word": "\u4e5f", "start": 8.3, "end": 8.44, "confidence": 0.6378956, "punctuated_word": "\u4e5f"}, {"word": "\u50b3", "start": 8.44, "end": 8.599999, "confidence": 0.34981716, "punctuated_word": "\u50b3"}, {"word": "\u56de\u4f86", "start": 8.599999, "end": 8.98, "confidence": 0.9960781, "punctuated_word": "\u56de\u4f86"}, {"word": "\u8a31", "start": 8.98, "end": 9.139999, "confidence": 0.9695547, "punctuated_word": "\u8a31"}, {"word": "\u591a", "start": 9.139999, "end": 9.28, "confidence": 0.9946556, "punctuated_word": "\u591a"}, {"word": "\u904e\u53bb", "start": 9.28, "end": 9.62, "confidence": 0.99605715, "punctuated_word": "\u904e\u53bb"}, {"word": "\u96e3", "start": 9.62, "end": 9.84, "confidence": 0.99637187, "punctuated_word": "\u96e3"}, {"word": "\u4ee5", "start": 9.84, "end": 9.96, "confidence": 0.9914835, "punctuated_word": "\u4ee5"}, {"word": "\u62cd", "start": 9.96, "end": 10.2, "confidence": 0.99999964, "punctuated_word": "\u62cd"}, {"word": "\u651d", "start": 10.2, "end": 10.36, "confidence": 0.98693615, "punctuated_word": "\u651d"}, {"word": "\u5230", "start": 10.36, "end": 10.5199995, "confidence": 0.99140453, "punctuated_word": "\u5230"}, {"word": "\u7684", "start": 10.5199995, "end": 10.639999, "confidence": 0.9991404, "punctuated_word": "\u7684"}, {"word": "\u7167", "start": 10.639999, "end": 10.84, "confidence": 0.09480816, "punctuated_word": "\u7167"}, {"word": "\u7247", "start": 10.84, "end": 11.34, "confidence": 0.9071409, "punctuated_word": "\u7247"}], "paragraphs": {"transcript": "\n\u53e4 \u8001 \u65b0 \u6232 \u4e2d \u767c\u73fe \u4e86 \u6709 \u6a5f \u5206 \u5b50 \u6211\u5011 \u96e2 \u7b2c\u4e09 \u985e \u63a5 \u89f8 \u9084\u6709 \u591a \u9060 \u5fae \u535a \u6b63 \u5f0f \u5c55 \u958b \u62cd \u651d \u4efb \u52d9 \u5df2\u7d93 \u5c46 \u6eff \u9031 \u5e74 \u6700\u8fd1 \u4e5f \u50b3 \u56de\u4f86 \u8a31 \u591a \u904e\u53bb \u96e3 \u4ee5 \u62cd \u651d \u5230 \u7684 \u7167 \u7247", "paragraphs": [{"sentences": [{"text": "\u53e4 \u8001 \u65b0 \u6232 \u4e2d \u767c\u73fe \u4e86 \u6709 \u6a5f \u5206 \u5b50 \u6211\u5011 \u96e2 \u7b2c\u4e09 \u985e \u63a5 \u89f8 \u9084\u6709 \u591a \u9060 \u5fae \u535a \u6b63 \u5f0f \u5c55 \u958b \u62cd \u651d \u4efb \u52d9 \u5df2\u7d93 \u5c46 \u6eff \u9031 \u5e74 \u6700\u8fd1 \u4e5f \u50b3 \u56de\u4f86 \u8a31 \u591a \u904e\u53bb \u96e3 \u4ee5 \u62cd \u651d \u5230 \u7684 \u7167 \u7247", "start": 0.06, "end": 11.34}], "start": 0.06, "end": 11.34, "num_words": 50}]}}]}], "utterances": [{"start": 0.06, "end": 11.34, "confidence": 0.84095657, "channel": 0, "transcript": "\u53e4 \u8001 \u65b0 \u6232 \u4e2d \u767c\u73fe \u4e86 \u6709 \u6a5f \u5206 \u5b50 \u6211\u5011 \u96e2 \u7b2c\u4e09 \u985e \u63a5 \u89f8 \u9084\u6709 \u591a \u9060 \u5fae \u535a \u6b63 \u5f0f \u5c55 \u958b \u62cd \u651d \u4efb \u52d9 \u5df2\u7d93 \u5c46 \u6eff \u9031 \u5e74 \u6700\u8fd1 \u4e5f \u50b3 \u56de\u4f86 \u8a31 \u591a \u904e\u53bb \u96e3 \u4ee5 \u62cd \u651d \u5230 \u7684 \u7167 \u7247", "words": [{"word": "\u53e4", "start": 0.06, "end": 0.22, "confidence": 0.24031715, "punctuated_word": "\u53e4"}, {"word": "\u8001", "start": 0.22, "end": 0.45999998, "confidence": 0.4851624, "punctuated_word": "\u8001"}, {"word": "\u65b0", "start": 0.45999998, "end": 0.52, "confidence": 0.9871694, "punctuated_word": "\u65b0"}, {"word": "\u6232", "start": 0.52, "end": 0.71999997, "confidence": 0.689627, "punctuated_word": "\u6232"}, {"word": "\u4e2d", "start": 0.71999997, "end": 0.94, "confidence": 0.7163843, "punctuated_word": "\u4e2d"}, {"word": "\u767c\u73fe", "start": 0.94, "end": 1.4, "confidence": 0.99306655, "punctuated_word": "\u767c\u73fe"}, {"word": "\u4e86", "start": 1.4, "end": 1.5, "confidence": 0.99730957, "punctuated_word": "\u4e86"}, {"word": "\u6709", "start": 1.5, "end": 1.7199999, "confidence": 0.9590886, "punctuated_word": "\u6709"}, {"word": "\u6a5f", "start": 1.7199999, "end": 1.88, "confidence": 0.99941957, "punctuated_word": "\u6a5f"}, {"word": "\u5206", "start": 1.88, "end": 2.08, "confidence": 0.055190008, "punctuated_word": "\u5206"}, {"word": "\u5b50", "start": 2.08, "end": 2.58, "confidence": 0.96733665, "punctuated_word": "\u5b50"}, {"word": "\u6211\u5011", "start": 2.6599998, "end": 2.8999999, "confidence": 0.9208613, "punctuated_word": "\u6211\u5011"}, {"word": "\u96e2", "start": 2.8999999, "end": 3.12, "confidence": 0.9793457, "punctuated_word": "\u96e2"}, {"word": "\u7b2c\u4e09", "start": 3.12, "end": 3.3799999, "confidence": 0.9081615, "punctuated_word": "\u7b2c\u4e09"}, {"word": "\u985e", "start": 3.3799999, "end": 3.6, "confidence": 0.99989676, "punctuated_word": "\u985e"}, {"word": "\u63a5", "start": 3.6, "end": 3.76, "confidence": 0.9999919, "punctuated_word": "\u63a5"}, {"word": "\u89f8", "start": 3.76, "end": 4.04, "confidence": 0.7521827, "punctuated_word": "\u89f8"}, {"word": "\u9084\u6709", "start": 4.04, "end": 4.3199997, "confidence": 0.9909378, "punctuated_word": "\u9084\u6709"}, {"word": "\u591a", "start": 4.3199997, "end": 4.48, "confidence": 0.17423469, "punctuated_word": "\u591a"}, {"word": "\u9060", "start": 4.48, "end": 4.98, "confidence": 0.9891718, "punctuated_word": "\u9060"}, {"word": "\u5fae", "start": 5.02, "end": 5.2, "confidence": 0.99999905, "punctuated_word": "\u5fae"}, {"word": "\u535a", "start": 5.2, "end": 5.38, "confidence": 0.9900187, "punctuated_word": "\u535a"}, {"word": "\u6b63", "start": 5.38, "end": 5.54, "confidence": 0.9524369, "punctuated_word": "\u6b63"}, {"word": "\u5f0f", "start": 5.54, "end": 5.68, "confidence": 0.9987935, "punctuated_word": "\u5f0f"}, {"word": "\u5c55", "start": 5.68, "end": 5.8599997, "confidence": 0.9839077, "punctuated_word": "\u5c55"}, {"word": "\u958b", "start": 5.8599997, "end": 6.04, "confidence": 0.9992906, "punctuated_word": "\u958b"}, {"word": "\u62cd", "start": 6.04, "end": 6.18, "confidence": 0.99999917, "punctuated_word": "\u62cd"}, {"word": "\u651d", "start": 6.18, "end": 6.3799996, "confidence": 0.99359775, "punctuated_word": "\u651d"}, {"word": "\u4efb", "start": 6.3799996, "end": 6.5, "confidence": 0.13850875, "punctuated_word": "\u4efb"}, {"word": "\u52d9", "start": 6.5, "end": 6.7599998, "confidence": 0.9938491, "punctuated_word": "\u52d9"}, {"word": "\u5df2\u7d93", "start": 6.7599998, "end": 6.96, "confidence": 0.9992866, "punctuated_word": "\u5df2\u7d93"}, {"word": "\u5c46", "start": 6.96, "end": 7.12, "confidence": 0.9860264, "punctuated_word": "\u5c46"}, {"word": "\u6eff", "start": 7.12, "end": 7.3599997, "confidence": 0.9997645, "punctuated_word": "\u6eff"}, {"word": "\u9031", "start": 7.3599997, "end": 7.54, "confidence": 0.33030534, "punctuated_word": "\u9031"}, {"word": "\u5e74", "start": 7.54, "end": 8.0, "confidence": 0.9914724, "punctuated_word": "\u5e74"}, {"word": "\u6700\u8fd1", "start": 8.0, "end": 8.3, "confidence": 0.97437614, "punctuated_word": "\u6700\u8fd1"}, {"word": "\u4e5f", "start": 8.3, "end": 8.44, "confidence": 0.6378956, "punctuated_word": "\u4e5f"}, {"word": "\u50b3", "start": 8.44, "end": 8.599999, "confidence": 0.34981716, "punctuated_word": "\u50b3"}, {"word": "\u56de\u4f86", "start": 8.599999, "end": 8.98, "confidence": 0.9960781, "punctuated_word": "\u56de\u4f86"}, {"word": "\u8a31", "start": 8.98, "end": 9.139999, "confidence": 0.9695547, "punctuated_word": "\u8a31"}, {"word": "\u591a", "start": 9.139999, "end": 9.28, "confidence": 0.9946556, "punctuated_word": "\u591a"}, {"word": "\u904e\u53bb", "start": 9.28, "end": 9.62, "confidence": 0.99605715, "punctuated_word": "\u904e\u53bb"}, {"word": "\u96e3", "start": 9.62, "end": 9.84, "confidence": 0.99637187, "punctuated_word": "\u96e3"}, {"word": "\u4ee5", "start": 9.84, "end": 9.96, "confidence": 0.9914835, "punctuated_word": "\u4ee5"}, {"word": "\u62cd", "start": 9.96, "end": 10.2, "confidence": 0.99999964, "punctuated_word": "\u62cd"}, {"word": "\u651d", "start": 10.2, "end": 10.36, "confidence": 0.98693615, "punctuated_word": "\u651d"}, {"word": "\u5230", "start": 10.36, "end": 10.5199995, "confidence": 0.99140453, "punctuated_word": "\u5230"}, {"word": "\u7684", "start": 10.5199995, "end": 10.639999, "confidence": 0.9991404, "punctuated_word": "\u7684"}, {"word": "\u7167", "start": 10.639999, "end": 10.84, "confidence": 0.09480816, "punctuated_word": "\u7167"}, {"word": "\u7247", "start": 10.84, "end": 11.34, "confidence": 0.9071409, "punctuated_word": "\u7247"}], "id": "bb514832-2d54-4c0a-beb4-17e7f539ea5f"}]}}

videotrans/__init__.py

@@ -1,4 +1,4 @@
 # -*- coding: utf-8 -*-
 
-VERSION = "v2.96"
-VERSION_NUM = 120096
+VERSION = "v2.97"
+VERSION_NUM = 120097

videotrans/component/__init__.py

@@ -1,10 +1,10 @@
 from videotrans.component.set_form import BaiduForm, ChatgptForm, DeepLForm, DeepLXForm, TencentForm, ElevenlabsForm, \
     InfoForm, AzureForm, GeminiForm, SetLineRole, YoutubeForm, OttForm, CloneForm, SeparateForm, TtsapiForm, \
-    GPTSoVITSForm, TransapiForm, ArticleForm, AzurettsForm, ZhrecognForm, ChatttsForm, LocalLLMForm, ZijiehuoshanForm, \
+    GPTSoVITSForm, TransapiForm, ArticleForm, AzurettsForm,  ChatttsForm, LocalLLMForm, ZijiehuoshanForm, \
     HebingsrtForm, DoubaoForm, FishTTSForm, CosyVoiceForm, AI302Form, AI302TTSForm, SetINIForm, WatermarkForm, \
     GetaudioForm, HunliuForm, VASForm, Fanyisrt, Recognform, Peiyinform, Videoandaudioform, Videoandsrtform, \
     OpenAITTSForm, RecognAPIForm, OpenaiRecognAPIForm, DownloadModelForm, FormatcoverForm, SubtitlescoverForm, \
-    SubtitleEditer,SttAPIForm,VolcEngineTTSForm,SenseVoiceAPIForm,F5TTSForm,DeepgramForm
+    SubtitleEditer,SttAPIForm,VolcEngineTTSForm,F5TTSForm,DeepgramForm
 
 __all__ = [
     "BaiduForm",
@@ -26,7 +26,6 @@
     "TransapiForm",
     "ArticleForm",
     "AzurettsForm",
-    "ZhrecognForm",
     "ChatttsForm",
     "LocalLLMForm",
     "ZijiehuoshanForm",
@@ -55,7 +54,6 @@
     "SubtitleEditer",
     "SttAPIForm",
     "VolcEngineTTSForm",
-    "SenseVoiceAPIForm",
     "F5TTSForm",
     "DeepgramForm"
 ]

videotrans/component/set_form.py

@@ -36,7 +36,6 @@
 from videotrans.ui.recogn import Ui_recogn
 from videotrans.ui.recognapi import Ui_recognapiform
 from videotrans.ui.stt import Ui_sttform
-from videotrans.ui.sense import Ui_senseform
 from videotrans.ui.separate import Ui_separateform
 from videotrans.ui.setini import Ui_setini
 from videotrans.ui.setlinerole import Ui_setlinerole
@@ -52,7 +51,6 @@
 from videotrans.ui.volcenginetts import Ui_volcengineform
 from videotrans.ui.watermark import Ui_watermark
 from videotrans.ui.youtube import Ui_youtubeform
-from videotrans.ui.zh_recogn import Ui_zhrecognform
 from videotrans.ui.zijiehuoshan import Ui_zijiehuoshanform
 from videotrans.ui.f5tts import Ui_f5ttsform
 
@@ -138,14 +136,6 @@ def __init__(self, parent=None):
         self.setWindowIcon(QIcon(f"{config.ROOT_DIR}/videotrans/styles/icon.ico"))
 
 
-class SenseVoiceAPIForm(QDialog, Ui_senseform):  # <===
-    def __init__(self, parent=None):
-        super(SenseVoiceAPIForm, self).__init__(parent)
-        self.setupUi(self)
-        self.setWindowIcon(QIcon(f"{config.ROOT_DIR}/videotrans/styles/icon.ico"))
-
-
-
 class OpenaiRecognAPIForm(QDialog, Ui_openairecognapiform):  # <===
     def __init__(self, parent=None):
         super(OpenaiRecognAPIForm, self).__init__(parent)
@@ -270,11 +260,6 @@ def __init__(self, parent=None):
         self.setWindowIcon(QIcon(f"{config.ROOT_DIR}/videotrans/styles/icon.ico"))
 
 
-class ZhrecognForm(QDialog, Ui_zhrecognform):  # <===
-    def __init__(self, parent=None):
-        super(ZhrecognForm, self).__init__(parent)
-        self.setupUi(self)
-        self.setWindowIcon(QIcon(f"{config.ROOT_DIR}/videotrans/styles/icon.ico"))
 
 
 class DoubaoForm(QDialog, Ui_doubaoform):  # <===

videotrans/configure/config.py

@@ -132,6 +132,7 @@ def _log_uncaught_exceptions(exc_type, exc_value, exc_traceback):
 os.environ['QT_API'] = 'pyside6'
 os.environ['SOFT_NAME'] = 'pyvideotrans'
 os.environ['MODELSCOPE_CACHE'] = ROOT_DIR + "/models"
+os.environ['HF_HOME'] = ROOT_DIR + "/models"
 ####################################
 # 存储所有任务的进度队列，以uuid为键
 # 根据uuid将日志进度等信息存入队列，如果不存在则创建
@@ -201,6 +202,20 @@ def push_queue(uuid, jsondata):
 # 执行模式 gui 或 api
 exec_mode="gui"
 
+# funasr模型
+FUNASR_MODEL=['paraformer-zh','SenseVoiceSmall']
+DEEPGRAM_MODEL=[
+            "whisper-large",
+            "whisper-medium",
+            "whisper-small",
+            "whisper-base",
+            "whisper-tiny",
+            "nova-2-general",
+            "enhanced-2-general",
+            "base-2-general",
+
+        ]
+
 
 # 支持的视频格式
 VIDEO_EXTS = ["mp4", "mkv", "mpeg", "avi", "mov"]
@@ -675,10 +690,9 @@ def getset_params(obj=None):
 
         "stt_source_language":0,
         "stt_recogn_type":0,
-        "stt_model_name":0,
+        "stt_model_name":"",
 
         "deepgram_apikey":"",
-        "deepgram_model":"whisper-large",
         "deepgram_utt":200,
 
         "trans_translate_type":0,