jonnatakusuma commited on
Commit
da1cda6
1 Parent(s): a477f7d
Files changed (1) hide show
  1. app.py +48 -48
app.py CHANGED
@@ -1,64 +1,64 @@
1
- # import streamlit as st
2
- # import whisper
3
- # from tempfile import NamedTemporaryFile
4
- # import ffmpeg
5
 
6
 
7
- # st.title("MinuteBot App")
8
 
9
- # # upload audio file with streamlit
10
- # audio_file = st.file_uploader("Unggah Meeting Audio", type=["mp3", "wav", "m4a"])
11
 
12
- # # model = whisper.load_model("base") # loading the base model
13
- # st.text("MinuteBot Model telah dimuat:")
14
 
15
- # def load_whisper_model():
16
 
17
- # return model
18
 
19
 
20
- # if st.sidebar.button("Transkripsikan Audio"):
21
- # if audio_file is not None:
22
- # with NamedTemporaryFile() as temp:
23
- # temp.write(audio_file.getvalue())
24
- # temp.seek(0)
25
- # model = whisper.load_model("large")
26
- # result = model.transcribe(temp.name)
27
- # st.write(result["text"])
28
 
29
- # st.sidebar.header("Putar Berkas Audio")
30
- # st.sidebar.audio(audio_file)
31
 
32
- import streamlit as st
33
- from tempfile import NamedTemporaryFile
34
- import ffmpeg
35
- from transformers import Wav2Vec2ForCTC, Wav2Vec2Tokenizer
36
- import librosa
37
 
38
- # HF_TOKEN = os.environ.get("HUGGINGFACEHUB_API_TOKEN")
39
 
40
- st.title("TemplarX-Medium-Indonesian Transcription App")
41
- st.text("Model Whisper (TemplarX-medium-Indonesian) telah dimuat:")
42
 
43
- def load_whisper_model():
44
- model_name = "jonnatakusuma/TemplarX-medium-Indonesian"
45
- tokenizer = Wav2Vec2Tokenizer.from_pretrained(model_name)
46
- model = Wav2Vec2ForCTC.from_pretrained(model_name, use_auth_token=True)
47
- return tokenizer, model
48
 
49
- audio_file = st.file_uploader("Unggah Meeting Audio", type=["mp3", "wav", "m4a"])
50
 
51
- if st.sidebar.button("Transkripsikan Audio"):
52
- if audio_file is not None:
53
- with NamedTemporaryFile() as temp:
54
- temp.write(audio_file.read())
55
- temp.seek(0)
56
- tokenizer, model = load_whisper_model()
57
- # Read the audio file and transcribe using the fine-tuned model
58
- audio_path = temp.name
59
- audio_input, _ = librosa.load(audio_path, sr=16000)
60
- transcription = model.stt(text)
61
- st.write(transcription)
62
 
63
- st.sidebar.header("Putar Berkas Audio")
64
- st.sidebar.audio(audio_file, format='audio/wav')
 
1
+ import streamlit as st
2
+ import whisper
3
+ from tempfile import NamedTemporaryFile
4
+ import ffmpeg
5
 
6
 
7
+ st.title("MinuteBot App")
8
 
9
+ # upload audio file with streamlit
10
+ audio_file = st.file_uploader("Unggah Meeting Audio", type=["mp3", "wav", "m4a"])
11
 
12
+ # model = whisper.load_model("base") # loading the base model
13
+ st.text("MinuteBot Model telah dimuat:")
14
 
15
+ def load_whisper_model():
16
 
17
+ return model
18
 
19
 
20
+ if st.sidebar.button("Transkripsikan Audio"):
21
+ if audio_file is not None:
22
+ with NamedTemporaryFile() as temp:
23
+ temp.write(audio_file.getvalue())
24
+ temp.seek(0)
25
+ model = whisper.load_model("large")
26
+ result = model.transcribe(temp.name)
27
+ st.write(result["text"])
28
 
29
+ st.sidebar.header("Putar Berkas Audio")
30
+ st.sidebar.audio(audio_file)
31
 
32
+ # import streamlit as st
33
+ # from tempfile import NamedTemporaryFile
34
+ # import ffmpeg
35
+ # from transformers import Wav2Vec2ForCTC, Wav2Vec2Tokenizer
36
+ # import librosa
37
 
38
+ # # HF_TOKEN = os.environ.get("HUGGINGFACEHUB_API_TOKEN")
39
 
40
+ # st.title("TemplarX-Medium-Indonesian Transcription App")
41
+ # st.text("Model Whisper (TemplarX-medium-Indonesian) telah dimuat:")
42
 
43
+ # def load_whisper_model():
44
+ # model_name = "jonnatakusuma/TemplarX-medium-Indonesian"
45
+ # tokenizer = Wav2Vec2Tokenizer.from_pretrained(model_name)
46
+ # model = Wav2Vec2ForCTC.from_pretrained(model_name, use_auth_token=True)
47
+ # return tokenizer, model
48
 
49
+ # audio_file = st.file_uploader("Unggah Meeting Audio", type=["mp3", "wav", "m4a"])
50
 
51
+ # if st.sidebar.button("Transkripsikan Audio"):
52
+ # if audio_file is not None:
53
+ # with NamedTemporaryFile() as temp:
54
+ # temp.write(audio_file.read())
55
+ # temp.seek(0)
56
+ # tokenizer, model = load_whisper_model()
57
+ # # Read the audio file and transcribe using the fine-tuned model
58
+ # audio_path = temp.name
59
+ # audio_input, _ = librosa.load(audio_path, sr=16000)
60
+ # transcription = model.stt(text)
61
+ # st.write(transcription)
62
 
63
+ # st.sidebar.header("Putar Berkas Audio")
64
+ # st.sidebar.audio(audio_file, format='audio/wav')