Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -2,12 +2,12 @@ import torch
|
|
| 2 |
from transformers import pipeline
|
| 3 |
import gradio as gr
|
| 4 |
|
| 5 |
-
# Configuraci贸n
|
| 6 |
MODEL_NAME = "openai/whisper-small"
|
| 7 |
BATCH_SIZE = 8
|
| 8 |
-
device = 0 if torch.cuda.is_available() else "cpu"
|
| 9 |
|
| 10 |
-
#
|
|
|
|
|
|
|
| 11 |
pipe = pipeline(
|
| 12 |
task="automatic-speech-recognition",
|
| 13 |
model=MODEL_NAME,
|
|
@@ -30,57 +30,48 @@ def format_timestamp(seconds: float, always_include_hours: bool = False, decimal
|
|
| 30 |
|
| 31 |
def transcribe(file, task, return_timestamps):
|
| 32 |
if file is None:
|
| 33 |
-
return "
|
| 34 |
-
|
| 35 |
-
outputs = pipe(
|
| 36 |
-
file,
|
| 37 |
-
batch_size=BATCH_SIZE,
|
| 38 |
-
generate_kwargs={"task": task},
|
| 39 |
-
return_timestamps=return_timestamps
|
| 40 |
-
)
|
| 41 |
-
|
| 42 |
text = outputs["text"]
|
|
|
|
| 43 |
if return_timestamps:
|
| 44 |
-
|
| 45 |
-
|
| 46 |
f"[{format_timestamp(chunk['timestamp'][0])} -> {format_timestamp(chunk['timestamp'][1])}] {chunk['text']}"
|
| 47 |
-
for chunk in
|
| 48 |
]
|
| 49 |
-
text = "\n".join(
|
| 50 |
return text
|
| 51 |
|
| 52 |
-
#
|
| 53 |
mic_transcribe = gr.Interface(
|
| 54 |
fn=transcribe,
|
| 55 |
inputs=[
|
| 56 |
-
gr.Audio(sources="microphone", type="filepath"
|
| 57 |
-
gr.Radio(["transcribe", "translate"], label="
|
| 58 |
-
gr.Checkbox(
|
| 59 |
],
|
| 60 |
outputs="text",
|
| 61 |
-
title="Whisper Demo:
|
| 62 |
-
|
| 63 |
-
allow_flagging="never",
|
| 64 |
)
|
| 65 |
|
| 66 |
-
# Interfaz de Archivo
|
| 67 |
file_transcribe = gr.Interface(
|
| 68 |
fn=transcribe,
|
| 69 |
inputs=[
|
| 70 |
-
gr.Audio(sources="upload", type="filepath", label="
|
| 71 |
-
gr.Radio(["transcribe", "translate"], label="
|
| 72 |
-
gr.Checkbox(
|
| 73 |
],
|
| 74 |
outputs="text",
|
| 75 |
-
title="Whisper Demo:
|
| 76 |
-
|
| 77 |
-
allow_flagging="never",
|
| 78 |
)
|
| 79 |
|
| 80 |
-
# Crear la aplicaci贸n con pesta帽as
|
| 81 |
demo = gr.TabbedInterface(
|
| 82 |
[mic_transcribe, file_transcribe],
|
| 83 |
-
["
|
| 84 |
)
|
| 85 |
|
| 86 |
if __name__ == "__main__":
|
|
|
|
| 2 |
from transformers import pipeline
|
| 3 |
import gradio as gr
|
| 4 |
|
|
|
|
| 5 |
MODEL_NAME = "openai/whisper-small"
|
| 6 |
BATCH_SIZE = 8
|
|
|
|
| 7 |
|
| 8 |
+
# Forzar uso de CPU
|
| 9 |
+
device = -1
|
| 10 |
+
|
| 11 |
pipe = pipeline(
|
| 12 |
task="automatic-speech-recognition",
|
| 13 |
model=MODEL_NAME,
|
|
|
|
| 30 |
|
| 31 |
def transcribe(file, task, return_timestamps):
|
| 32 |
if file is None:
|
| 33 |
+
return "Error: No se proporcion贸 archivo de audio."
|
| 34 |
+
|
| 35 |
+
outputs = pipe(file, batch_size=BATCH_SIZE, generate_kwargs={"task": task}, return_timestamps=return_timestamps)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 36 |
text = outputs["text"]
|
| 37 |
+
|
| 38 |
if return_timestamps:
|
| 39 |
+
timestamps = outputs["chunks"]
|
| 40 |
+
timestamps = [
|
| 41 |
f"[{format_timestamp(chunk['timestamp'][0])} -> {format_timestamp(chunk['timestamp'][1])}] {chunk['text']}"
|
| 42 |
+
for chunk in timestamps
|
| 43 |
]
|
| 44 |
+
text = "\n".join(timestamps)
|
| 45 |
return text
|
| 46 |
|
| 47 |
+
# Configuraci贸n de interfaces con sintaxis moderna (Gradio 4+)
|
| 48 |
mic_transcribe = gr.Interface(
|
| 49 |
fn=transcribe,
|
| 50 |
inputs=[
|
| 51 |
+
gr.Audio(sources="microphone", type="filepath"),
|
| 52 |
+
gr.Radio(["transcribe", "translate"], label="Task", value="transcribe"),
|
| 53 |
+
gr.Checkbox(label="Return timestamps", value=False),
|
| 54 |
],
|
| 55 |
outputs="text",
|
| 56 |
+
title="Whisper Demo: Transcribe Microphone",
|
| 57 |
+
flagging_mode="never", # 'allow_flagging' ahora es 'flagging_mode'
|
|
|
|
| 58 |
)
|
| 59 |
|
|
|
|
| 60 |
file_transcribe = gr.Interface(
|
| 61 |
fn=transcribe,
|
| 62 |
inputs=[
|
| 63 |
+
gr.Audio(sources="upload", type="filepath", label="Audio file"),
|
| 64 |
+
gr.Radio(["transcribe", "translate"], label="Task", value="transcribe"),
|
| 65 |
+
gr.Checkbox(label="Return timestamps", value=False),
|
| 66 |
],
|
| 67 |
outputs="text",
|
| 68 |
+
title="Whisper Demo: Transcribe Audio File",
|
| 69 |
+
flagging_mode="never",
|
|
|
|
| 70 |
)
|
| 71 |
|
|
|
|
| 72 |
demo = gr.TabbedInterface(
|
| 73 |
[mic_transcribe, file_transcribe],
|
| 74 |
+
["Transcribe Microphone", "Transcribe Audio File"]
|
| 75 |
)
|
| 76 |
|
| 77 |
if __name__ == "__main__":
|