rebego commited on
Commit
4760e30
·
verified ·
1 Parent(s): e7977e9

Use whisper-small and remove source param

Browse files
Files changed (1) hide show
  1. app.py +6 -7
app.py CHANGED
@@ -2,27 +2,26 @@ import gradio as gr
2
  import torch
3
  from transformers import pipeline
4
 
5
- # 1) Pipeline de Whisper para ES → texto ES
6
  device = 0 if torch.cuda.is_available() else -1
7
  asr = pipeline(
8
  "automatic-speech-recognition",
9
- model="openai/whisper-medium",
10
  device=device,
11
- generate_kwargs={"task":"transcribe","language":"es"}
12
  )
13
 
14
  # 2) Función de transcripción
15
  def transcribe(audio_path):
16
- result = asr(audio_path)
17
- return result["text"]
18
 
19
  # 3) Interfaz Gradio
20
  demo = gr.Interface(
21
  fn=transcribe,
22
- inputs=gr.Audio(source="upload", type="filepath", label="Sube audio (ES)"),
23
  outputs=gr.Textbox(label="Transcripción"),
24
  title="Audio→Texto en Español",
25
- description="Transcribe audio en español con Whisper"
26
  )
27
 
28
  if __name__ == "__main__":
 
2
  import torch
3
  from transformers import pipeline
4
 
5
+ # 1) Pipeline de Whisper-small para ES → texto ES
6
  device = 0 if torch.cuda.is_available() else -1
7
  asr = pipeline(
8
  "automatic-speech-recognition",
9
+ model="openai/whisper-small", # <-- modelo pequeño para CPU
10
  device=device,
11
+ generate_kwargs={"task": "transcribe", "language": "es"}
12
  )
13
 
14
  # 2) Función de transcripción
15
  def transcribe(audio_path):
16
+ return asr(audio_path)["text"]
 
17
 
18
  # 3) Interfaz Gradio
19
  demo = gr.Interface(
20
  fn=transcribe,
21
+ inputs=gr.Audio(type="filepath", label="Sube audio (ES)"), # sin source="upload"
22
  outputs=gr.Textbox(label="Transcripción"),
23
  title="Audio→Texto en Español",
24
+ description="Transcribe audio en español con Whisper-small"
25
  )
26
 
27
  if __name__ == "__main__":