Spaces:
Sleeping
Sleeping
Merge branch 'eason/main' of https://github.com/project-kxkg/project-t into eason/main
Browse files- pipeline.py +3 -3
pipeline.py
CHANGED
@@ -98,21 +98,21 @@ def get_srt_class(srt_file_en, result_path, video_name, audio_path, audio_file =
|
|
98 |
# using whisper to perform speech-to-text and save it in <video name>_en.txt under RESULT PATH.
|
99 |
srt_file_en = "{}/{}/{}_en.srt".format(result_path, video_name, video_name)
|
100 |
if not os.path.exists(srt_file_en):
|
101 |
-
|
|
|
102 |
# use OpenAI API for transcribe
|
103 |
if method == "api":
|
104 |
transcript = openai.Audio.transcribe("whisper-1", audio_file)
|
105 |
|
106 |
# use local whisper model
|
107 |
elif method == "basic":
|
108 |
-
model = whisper.load_model(whisper_model) # using base model in local machine (may use large model on our server)
|
109 |
transcript = model.transcribe(audio_path)
|
110 |
|
111 |
# use stable-whisper
|
112 |
elif method == "stable":
|
113 |
|
114 |
# use cuda if available
|
115 |
-
devices = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
|
116 |
model = stable_whisper.load_model(whisper_model, device = devices)
|
117 |
transcript = model.transcribe(audio_path, regroup = False, initial_prompt="Hello, welcome to my lecture. Are you good my friend?")
|
118 |
(
|
|
|
98 |
# using whisper to perform speech-to-text and save it in <video name>_en.txt under RESULT PATH.
|
99 |
srt_file_en = "{}/{}/{}_en.srt".format(result_path, video_name, video_name)
|
100 |
if not os.path.exists(srt_file_en):
|
101 |
+
|
102 |
+
devices = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
|
103 |
# use OpenAI API for transcribe
|
104 |
if method == "api":
|
105 |
transcript = openai.Audio.transcribe("whisper-1", audio_file)
|
106 |
|
107 |
# use local whisper model
|
108 |
elif method == "basic":
|
109 |
+
model = whisper.load_model(whisper_model, device = devices) # using base model in local machine (may use large model on our server)
|
110 |
transcript = model.transcribe(audio_path)
|
111 |
|
112 |
# use stable-whisper
|
113 |
elif method == "stable":
|
114 |
|
115 |
# use cuda if available
|
|
|
116 |
model = stable_whisper.load_model(whisper_model, device = devices)
|
117 |
transcript = model.transcribe(audio_path, regroup = False, initial_prompt="Hello, welcome to my lecture. Are you good my friend?")
|
118 |
(
|