Spaces:
Running
Running
jhj0517
commited on
Commit
•
91b9b83
1
Parent(s):
736206b
better read
Browse files- modules/model_Inference.py +72 -66
modules/model_Inference.py
CHANGED
@@ -1,48 +1,52 @@
|
|
1 |
import whisper
|
2 |
-
from modules.subtitle_manager import get_srt,get_vtt,write_srt,write_vtt,safe_filename
|
3 |
-
from modules.youtube_manager import get_ytdata,get_ytaudio
|
4 |
import gradio as gr
|
5 |
import os
|
6 |
from datetime import datetime
|
7 |
|
8 |
-
DEFAULT_MODEL_SIZE="large-v2"
|
9 |
|
10 |
-
|
|
|
11 |
def __init__(self):
|
12 |
print("\nInitializing Model..\n")
|
13 |
self.current_model_size = DEFAULT_MODEL_SIZE
|
14 |
-
self.model = whisper.load_model(name=DEFAULT_MODEL_SIZE,download_root="models")
|
15 |
self.available_models = whisper.available_models()
|
16 |
self.available_langs = sorted(list(whisper.tokenizer.LANGUAGES.values()))
|
17 |
|
18 |
-
def transcribe_file(self,fileobjs
|
19 |
-
,model_size,lang,subformat,istranslate,
|
20 |
progress=gr.Progress()):
|
21 |
-
|
22 |
def progress_callback(progress_value):
|
23 |
-
progress(progress_value,desc="Transcribing..")
|
24 |
-
|
25 |
if model_size != self.current_model_size:
|
26 |
-
progress(0,desc="Initializing Model..")
|
27 |
self.current_model_size = model_size
|
28 |
-
self.model = whisper.load_model(name=model_size,download_root="models")
|
29 |
|
30 |
-
if lang == "Automatic Detection"
|
31 |
-
lang = None
|
32 |
|
33 |
-
progress(0,desc="Loading Audio..")
|
34 |
|
35 |
files_info = {}
|
36 |
-
for fileobj in fileobjs:
|
|
|
37 |
audio = whisper.load_audio(fileobj.name)
|
38 |
|
39 |
-
translatable_model = ["large","large-v1","large-v2"]
|
40 |
if istranslate and self.current_model_size in translatable_model:
|
41 |
-
result = self.model.transcribe(audio=audio,language=lang,verbose=False,task="translate",
|
42 |
-
|
43 |
-
|
|
|
|
|
44 |
|
45 |
-
progress(1,desc="Completed!")
|
46 |
|
47 |
file_name, file_ext = os.path.splitext(os.path.basename(fileobj.orig_name))
|
48 |
file_name = file_name[:-9]
|
@@ -52,47 +56,49 @@ class WhisperInference():
|
|
52 |
|
53 |
if subformat == "SRT":
|
54 |
subtitle = get_srt(result["segments"])
|
55 |
-
write_srt(subtitle,f"{output_path}.srt")
|
56 |
elif subformat == "WebVTT":
|
57 |
subtitle = get_vtt(result["segments"])
|
58 |
-
write_vtt(subtitle,f"{output_path}.vtt")
|
59 |
|
60 |
files_info[file_name] = subtitle
|
61 |
|
62 |
total_result = ''
|
63 |
-
for file_name,subtitle in files_info.items():
|
64 |
-
total_result+='------------------------------------\n'
|
65 |
-
total_result+=f'{file_name}\n\n'
|
66 |
-
total_result+=f'{subtitle}'
|
67 |
|
68 |
return f"Done! Subtitle is in the outputs folder.\n\n{total_result}"
|
69 |
-
|
70 |
-
def transcribe_youtube(self,youtubelink
|
71 |
-
|
72 |
-
|
73 |
-
|
74 |
def progress_callback(progress_value):
|
75 |
-
progress(progress_value,desc="Transcribing..")
|
76 |
|
77 |
if model_size != self.current_model_size:
|
78 |
-
progress(0,desc="Initializing Model..")
|
79 |
self.current_model_size = model_size
|
80 |
-
self.model = whisper.load_model(name=model_size,download_root="models")
|
81 |
|
82 |
-
if lang == "Automatic Detection"
|
83 |
-
lang = None
|
84 |
|
85 |
-
progress(0,desc="Loading Audio from Youtube..")
|
86 |
yt = get_ytdata(youtubelink)
|
87 |
audio = whisper.load_audio(get_ytaudio(yt))
|
88 |
|
89 |
-
translatable_model = ["large","large-v1","large-v2"]
|
90 |
if istranslate and self.current_model_size in translatable_model:
|
91 |
-
result = self.model.transcribe(audio=audio,language=lang,verbose=False,task="translate",
|
92 |
-
|
93 |
-
|
|
|
|
|
94 |
|
95 |
-
progress(1,desc="Completed!")
|
96 |
|
97 |
file_name = safe_filename(yt.title)
|
98 |
timestamp = datetime.now().strftime("%m%d%H%M%S")
|
@@ -100,48 +106,48 @@ class WhisperInference():
|
|
100 |
|
101 |
if subformat == "SRT":
|
102 |
subtitle = get_srt(result["segments"])
|
103 |
-
write_srt(subtitle,f"{output_path}.srt")
|
104 |
elif subformat == "WebVTT":
|
105 |
subtitle = get_vtt(result["segments"])
|
106 |
-
write_vtt(subtitle,f"{output_path}.vtt")
|
107 |
|
108 |
return f"Done! Subtitle file is in the outputs folder.\n\n{subtitle}"
|
109 |
-
|
110 |
-
def transcribe_mic(self,micaudio
|
111 |
-
|
112 |
-
|
113 |
|
114 |
def progress_callback(progress_value):
|
115 |
-
progress(progress_value,desc="Transcribing..")
|
116 |
-
|
117 |
if model_size != self.current_model_size:
|
118 |
-
progress(0,desc="Initializing Model..")
|
119 |
self.current_model_size = model_size
|
120 |
-
self.model = whisper.load_model(name=model_size,download_root="models")
|
121 |
|
122 |
-
if lang == "Automatic Detection"
|
123 |
-
lang = None
|
124 |
|
125 |
-
progress(0,desc="Loading Audio..")
|
126 |
|
127 |
-
translatable_model = ["large","large-v1","large-v2"]
|
128 |
if istranslate and self.current_model_size in translatable_model:
|
129 |
-
result = self.model.transcribe(audio=micaudio,language=lang,verbose=False,task="translate",
|
130 |
-
|
131 |
-
|
|
|
|
|
132 |
|
133 |
-
progress(1,desc="Completed!")
|
134 |
|
135 |
timestamp = datetime.now().strftime("%m%d%H%M%S")
|
136 |
output_path = f"outputs/Mic-{timestamp}"
|
137 |
|
138 |
if subformat == "SRT":
|
139 |
subtitle = get_srt(result["segments"])
|
140 |
-
write_srt(subtitle,f"{output_path}.srt")
|
141 |
elif subformat == "WebVTT":
|
142 |
subtitle = get_vtt(result["segments"])
|
143 |
-
write_vtt(subtitle,f"{output_path}.vtt")
|
144 |
-
|
145 |
return f"Done! Subtitle file is in the outputs folder.\n\n{subtitle}"
|
146 |
-
|
147 |
-
|
|
|
1 |
import whisper
|
2 |
+
from modules.subtitle_manager import get_srt, get_vtt, write_srt, write_vtt, safe_filename
|
3 |
+
from modules.youtube_manager import get_ytdata, get_ytaudio
|
4 |
import gradio as gr
|
5 |
import os
|
6 |
from datetime import datetime
|
7 |
|
8 |
+
DEFAULT_MODEL_SIZE = "large-v2"
|
9 |
|
10 |
+
|
11 |
+
class WhisperInference:
|
12 |
def __init__(self):
|
13 |
print("\nInitializing Model..\n")
|
14 |
self.current_model_size = DEFAULT_MODEL_SIZE
|
15 |
+
self.model = whisper.load_model(name=DEFAULT_MODEL_SIZE, download_root="models")
|
16 |
self.available_models = whisper.available_models()
|
17 |
self.available_langs = sorted(list(whisper.tokenizer.LANGUAGES.values()))
|
18 |
|
19 |
+
def transcribe_file(self, fileobjs
|
20 |
+
, model_size, lang, subformat, istranslate,
|
21 |
progress=gr.Progress()):
|
22 |
+
|
23 |
def progress_callback(progress_value):
|
24 |
+
progress(progress_value, desc="Transcribing..")
|
25 |
+
|
26 |
if model_size != self.current_model_size:
|
27 |
+
progress(0, desc="Initializing Model..")
|
28 |
self.current_model_size = model_size
|
29 |
+
self.model = whisper.load_model(name=model_size, download_root="models")
|
30 |
|
31 |
+
if lang == "Automatic Detection":
|
32 |
+
lang = None
|
33 |
|
34 |
+
progress(0, desc="Loading Audio..")
|
35 |
|
36 |
files_info = {}
|
37 |
+
for fileobj in fileobjs:
|
38 |
+
|
39 |
audio = whisper.load_audio(fileobj.name)
|
40 |
|
41 |
+
translatable_model = ["large", "large-v1", "large-v2"]
|
42 |
if istranslate and self.current_model_size in translatable_model:
|
43 |
+
result = self.model.transcribe(audio=audio, language=lang, verbose=False, task="translate",
|
44 |
+
progress_callback=progress_callback)
|
45 |
+
else:
|
46 |
+
result = self.model.transcribe(audio=audio, language=lang, verbose=False,
|
47 |
+
progress_callback=progress_callback)
|
48 |
|
49 |
+
progress(1, desc="Completed!")
|
50 |
|
51 |
file_name, file_ext = os.path.splitext(os.path.basename(fileobj.orig_name))
|
52 |
file_name = file_name[:-9]
|
|
|
56 |
|
57 |
if subformat == "SRT":
|
58 |
subtitle = get_srt(result["segments"])
|
59 |
+
write_srt(subtitle, f"{output_path}.srt")
|
60 |
elif subformat == "WebVTT":
|
61 |
subtitle = get_vtt(result["segments"])
|
62 |
+
write_vtt(subtitle, f"{output_path}.vtt")
|
63 |
|
64 |
files_info[file_name] = subtitle
|
65 |
|
66 |
total_result = ''
|
67 |
+
for file_name, subtitle in files_info.items():
|
68 |
+
total_result += '------------------------------------\n'
|
69 |
+
total_result += f'{file_name}\n\n'
|
70 |
+
total_result += f'{subtitle}'
|
71 |
|
72 |
return f"Done! Subtitle is in the outputs folder.\n\n{total_result}"
|
73 |
+
|
74 |
+
def transcribe_youtube(self, youtubelink
|
75 |
+
, model_size, lang, subformat, istranslate,
|
76 |
+
progress=gr.Progress()):
|
77 |
+
|
78 |
def progress_callback(progress_value):
|
79 |
+
progress(progress_value, desc="Transcribing..")
|
80 |
|
81 |
if model_size != self.current_model_size:
|
82 |
+
progress(0, desc="Initializing Model..")
|
83 |
self.current_model_size = model_size
|
84 |
+
self.model = whisper.load_model(name=model_size, download_root="models")
|
85 |
|
86 |
+
if lang == "Automatic Detection":
|
87 |
+
lang = None
|
88 |
|
89 |
+
progress(0, desc="Loading Audio from Youtube..")
|
90 |
yt = get_ytdata(youtubelink)
|
91 |
audio = whisper.load_audio(get_ytaudio(yt))
|
92 |
|
93 |
+
translatable_model = ["large", "large-v1", "large-v2"]
|
94 |
if istranslate and self.current_model_size in translatable_model:
|
95 |
+
result = self.model.transcribe(audio=audio, language=lang, verbose=False, task="translate",
|
96 |
+
progress_callback=progress_callback)
|
97 |
+
else:
|
98 |
+
result = self.model.transcribe(audio=audio, language=lang, verbose=False,
|
99 |
+
progress_callback=progress_callback)
|
100 |
|
101 |
+
progress(1, desc="Completed!")
|
102 |
|
103 |
file_name = safe_filename(yt.title)
|
104 |
timestamp = datetime.now().strftime("%m%d%H%M%S")
|
|
|
106 |
|
107 |
if subformat == "SRT":
|
108 |
subtitle = get_srt(result["segments"])
|
109 |
+
write_srt(subtitle, f"{output_path}.srt")
|
110 |
elif subformat == "WebVTT":
|
111 |
subtitle = get_vtt(result["segments"])
|
112 |
+
write_vtt(subtitle, f"{output_path}.vtt")
|
113 |
|
114 |
return f"Done! Subtitle file is in the outputs folder.\n\n{subtitle}"
|
115 |
+
|
116 |
+
def transcribe_mic(self, micaudio
|
117 |
+
, model_size, lang, subformat, istranslate,
|
118 |
+
progress=gr.Progress()):
|
119 |
|
120 |
def progress_callback(progress_value):
|
121 |
+
progress(progress_value, desc="Transcribing..")
|
122 |
+
|
123 |
if model_size != self.current_model_size:
|
124 |
+
progress(0, desc="Initializing Model..")
|
125 |
self.current_model_size = model_size
|
126 |
+
self.model = whisper.load_model(name=model_size, download_root="models")
|
127 |
|
128 |
+
if lang == "Automatic Detection":
|
129 |
+
lang = None
|
130 |
|
131 |
+
progress(0, desc="Loading Audio..")
|
132 |
|
133 |
+
translatable_model = ["large", "large-v1", "large-v2"]
|
134 |
if istranslate and self.current_model_size in translatable_model:
|
135 |
+
result = self.model.transcribe(audio=micaudio, language=lang, verbose=False, task="translate",
|
136 |
+
progress_callback=progress_callback)
|
137 |
+
else:
|
138 |
+
result = self.model.transcribe(audio=micaudio, language=lang, verbose=False,
|
139 |
+
progress_callback=progress_callback)
|
140 |
|
141 |
+
progress(1, desc="Completed!")
|
142 |
|
143 |
timestamp = datetime.now().strftime("%m%d%H%M%S")
|
144 |
output_path = f"outputs/Mic-{timestamp}"
|
145 |
|
146 |
if subformat == "SRT":
|
147 |
subtitle = get_srt(result["segments"])
|
148 |
+
write_srt(subtitle, f"{output_path}.srt")
|
149 |
elif subformat == "WebVTT":
|
150 |
subtitle = get_vtt(result["segments"])
|
151 |
+
write_vtt(subtitle, f"{output_path}.vtt")
|
152 |
+
|
153 |
return f"Done! Subtitle file is in the outputs folder.\n\n{subtitle}"
|
|
|
|