textToSQL commited on
Commit
4579397
·
1 Parent(s): 869e886

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +16 -11
app.py CHANGED
@@ -19,10 +19,23 @@ def get_completion(prompt, model='gpt-3.5-turbo'):
19
  return response.choices[0].message['content']
20
 
21
  def transcribe(audio):
22
- model = whisper.load_model("medium")
23
- result = model.transcribe(audio)
24
- return result.text
 
 
 
 
 
 
 
 
 
25
 
 
 
 
 
26
 
27
 
28
 
@@ -38,14 +51,6 @@ def transcribe(audio):
38
  # live=True).launch()
39
 
40
 
41
- def speech_to_text(speech):
42
- text = asr(speech)["text"]
43
- return text
44
-
45
-
46
- def text_to_sentiment(text):
47
- return classifier(text)[0]["label"]
48
-
49
 
50
  demo = gr.Blocks()
51
 
 
19
  return response.choices[0].message['content']
20
 
21
  def transcribe(audio):
22
+
23
+ #time.sleep(3)
24
+ # load audio and pad/trim it to fit 30 seconds
25
+ audio = whisper.load_audio(audio)
26
+ audio = whisper.pad_or_trim(audio)
27
+
28
+ # make log-Mel spectrogram and move to the same device as the model
29
+ mel = whisper.log_mel_spectrogram(audio).to(model.device)
30
+
31
+ # detect the spoken language
32
+ _, probs = model.detect_language(mel)
33
+ print(f"Detected language: {max(probs, key=probs.get)}")
34
 
35
+ # decode the audio
36
+ options = whisper.DecodingOptions(fp16 = False)
37
+ result = whisper.decode(model, mel, options)
38
+ return result.text
39
 
40
 
41
 
 
51
  # live=True).launch()
52
 
53
 
 
 
 
 
 
 
 
 
54
 
55
  demo = gr.Blocks()
56