0.6.0

CheshireCC · Mar 3, 2024 · f22de8b · f22de8b
1 parent 0531017
commit f22de8b
Show file tree

Hide file tree

Showing 8 changed files with 41 additions and 22 deletions.
diff --git a/fasterWhisperGUIConfig.json b/fasterWhisperGUIConfig.json
@@ -8,15 +8,15 @@
     "model_param": {
         "localModel": true,
         "onlineModel": false,
-        "model_path": "",
+        "model_path": "F:/WhisperModels/faster-whisper/whisper-tiny",
         "modelName": 0,
         "use_v3_model": false,
         "device": 1,
         "deviceIndex": "0",
-        "preciese": 5,
+        "preciese": 0,
         "thread_num": "4",
         "num_worker": "1",
-        "download_root": "",
+        "download_root": "C:/Users/12059/.cache/huggingface/hub",
         "local_files_only": false
     },
     "vad_param": {
@@ -41,10 +41,10 @@
         "language": 0,
         "task": false,
         "beam_size": "5",
-        "best_of": "1",
+        "best_of": "6",
         "patience": "1.0",
         "length_penalty": "1.0",
-        "temperature": "0",
+        "temperature": "0.0,0.2,0.4,0.6,0.8,1.0",
         "compression_ratio_threshold": "2.4",
         "log_prob_threshold": "-1.0",
         "no_speech_threshold": "0.6",
@@ -55,7 +55,7 @@
         "suppress_tokens": "-1",
         "without_timestamps": false,
         "max_initial_timestamp": "1.0",
-        "word_timestamps": false,
+        "word_timestamps": true,
         "prepend_punctuations": "\"'“¿([{-",
         "append_punctuations": "\"'.。,，!！?？:：”)]}、",
         "repetition_penalty": "1.0",
@@ -66,10 +66,10 @@
         "tabMovable": false,
         "tabScrollable": false,
         "tabShadowEnabled": false,
-        "tabMaxWidth": 200,
+        "tabMaxWidth": 220,
         "closeDisplayMode": 0,
-        "whisperXMinSpeaker": 0,
-        "whisperXMaxSpeaker": 0,
+        "whisperXMinSpeaker": 2,
+        "whisperXMaxSpeaker": 2,
         "outputFormat": 0,
         "outputEncoding": 1
     }

diff --git a/faster_whisper_GUI/aboutPageNavigationInterface.py b/faster_whisper_GUI/aboutPageNavigationInterface.py
@@ -112,7 +112,7 @@ def setupUI(self):
         self.mainVLayout.addWidget(HorizontalSeparator(self))
 
         self.strongBodyLabel = TitleLabel(self)
-        self.strongBodyLabel.setText(self.tr("协议"))
+        self.strongBodyLabel.setText(self.tr("生成式人工智能程序用户协议"))
         self.strongBodyLabel.setAlignment(Qt.AlignmentFlag.AlignCenter)
         self.addWidget(self.strongBodyLabel)
 

diff --git a/faster_whisper_GUI/config.py b/faster_whisper_GUI/config.py
@@ -114,8 +114,22 @@
                 'bfloat16'
             ]
 
-Model_names = ["tiny", "tiny.en", "base", "base.en", "small", 
-                "small.en", "medium", "medium.en", "large-v1", "large-v2","large-v3"]
+Model_names = [
+                "tiny", 
+                "tiny.en", 
+                "base", 
+                "base.en", 
+                "small", 
+                "small.en", 
+                "medium", 
+                "medium.en", 
+                "large-v1", 
+                "large-v2",
+                "large-v3",
+                "distil-large-v2",
+                "distil-medium.en",
+                "distil-small.en"
+            ]
 
 Device_list = ["cpu", "cuda", "auto"]
 Task_list = ["transcribe" , "translate"]
@@ -157,6 +171,7 @@
 
 THEME_COLORS = [
     "#009faa",
+    "#81D8CF",
     "#ff009f",
     "#84BE84",
     "#aaff00",

diff --git a/faster_whisper_GUI/mainWindows.py b/faster_whisper_GUI/mainWindows.py
@@ -1015,7 +1015,7 @@ def raiseSuccessInfoBar(self, title:str, content:str):
                         title=title
                         , content=content
                         , isClosable=True
-                        , duration=-1
+                        , duration=5000
                         , position=InfoBarPosition.TOP
                         , parent=self
                     )
@@ -1447,14 +1447,14 @@ def deleteResultTableEvent(self, routeKey:str):
         self.outputWithDateTime("deleteTable")
 
         print(f"len_DataModel:{len(self.tableModel_list)}")
-        print(routeKey)
+        # print(routeKey)
         file_key ="_".join(routeKey.split("_")[1:]) 
         print(file_key)
         self.tableModel_list.pop(file_key)
         print(f"len_DataModel_after_pop:{len(self.tableModel_list)}")
 
         for result in [se for se in [self.current_result ,self.result_faster_whisper, self.result_whisperx_aligment, self.result_whisperx_speaker_diarize] if se is not None]:
-            print(f"len_result: {result}")
+            print(f"len_result: {len(result)}")
             for segmengs in result:
                 if segmengs[1] == file_key:
                     result.remove(segmengs)

diff --git a/faster_whisper_GUI/transcribe.py b/faster_whisper_GUI/transcribe.py
@@ -5,11 +5,8 @@
 import os
 from typing import List
 import time
-
 import codecs
-
 import torch
-
 import numpy as np 
 import av
 

diff --git a/faster_whisper_GUI/version.py b/faster_whisper_GUI/version.py
@@ -1,6 +1,6 @@
 # coding:utf-8
 
-__version__ = "0.5.7"
-__FasterWhisper_version__ = "0.10.0"
+__version__ = "0.6.0"
+__FasterWhisper_version__ = "1.0.1"
 __WhisperX_version__ = "3.1.1"
 __Demucs_version__ = "v4.0"
diff --git a/huggingface-config.json b/huggingface-config.json
@@ -0,0 +1,6 @@
+{
+    "proxy": {
+                "http": "http://127.0.0.1:7890",
+                "https": "https://127.0.0.1:7890"
+            }
+}
diff --git a/whisperx/alignment.py b/whisperx/alignment.py
@@ -19,7 +19,7 @@
 
 PUNKT_ABBREVIATIONS = ['dr', 'vs', 'mr', 'mrs', 'prof']
 
-LANGUAGES_WITHOUT_SPACES = ["ja", "zh"]
+LANGUAGES_WITHOUT_SPACES = ["ja", "zh", "yue"]
 
 DEFAULT_ALIGN_MODELS_TORCH = {
     "en": "WAV2VEC2_ASR_BASE_960H",
@@ -50,7 +50,8 @@
     "ko": "kresnik/wav2vec2-large-xlsr-korean",
     "ur": "kingabzpro/wav2vec2-large-xls-r-300m-Urdu",
     "te": "anuragshas/wav2vec2-large-xlsr-53-telugu",
-    "hi": "theainerd/Wav2Vec2-large-xlsr-hindi"
+    "hi": "theainerd/Wav2Vec2-large-xlsr-hindi",
+    "yue":"ctl/wav2vec2-large-xlsr-cantonese"
 }