Spaces:
Paused
Paused
Update app.py
Browse filessecond test Blocks()
app.py
CHANGED
|
@@ -7,15 +7,13 @@ import ffmpeg # Make sure it's ffmpeg-python
|
|
| 7 |
# Check if GPU is available
|
| 8 |
use_gpu = torch.cuda.is_available()
|
| 9 |
|
| 10 |
-
|
| 11 |
# Configure the pipeline to use the GPU if available
|
| 12 |
if use_gpu:
|
| 13 |
-
p = pipeline("automatic-speech-recognition",
|
| 14 |
-
|
| 15 |
else:
|
| 16 |
-
p = pipeline("automatic-speech-recognition",
|
| 17 |
-
|
| 18 |
-
|
| 19 |
|
| 20 |
def extract_audio_from_m3u8(url):
|
| 21 |
try:
|
|
@@ -25,8 +23,7 @@ def extract_audio_from_m3u8(url):
|
|
| 25 |
except Exception as e:
|
| 26 |
return f"An error occurred: {e}"
|
| 27 |
|
| 28 |
-
|
| 29 |
-
def transcribe(audio, state="", uploaded_audio=None, m3u8_url=""):
|
| 30 |
if m3u8_url:
|
| 31 |
audio = extract_audio_from_m3u8(m3u8_url)
|
| 32 |
|
|
@@ -34,36 +31,48 @@ def transcribe(audio, state="", uploaded_audio=None, m3u8_url=""):
|
|
| 34 |
audio = uploaded_audio
|
| 35 |
|
| 36 |
if not audio:
|
| 37 |
-
return state, state # Return a meaningful message
|
| 38 |
|
| 39 |
try:
|
| 40 |
time.sleep(3)
|
| 41 |
text = p(audio, chunk_length_s= 50)["text"]
|
| 42 |
state += text + "\n"
|
| 43 |
-
return state, state
|
| 44 |
except Exception as e:
|
| 45 |
-
return "An error occurred during transcription.", state # Handle other exceptions
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 46 |
|
|
|
|
|
|
|
| 47 |
|
| 48 |
-
|
| 49 |
-
|
| 50 |
-
|
|
|
|
|
|
|
| 51 |
|
|
|
|
|
|
|
| 52 |
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
|
| 56 |
-
gr.components.Audio(source="microphone", type="filepath"),
|
| 57 |
-
'state',
|
| 58 |
-
gr.components.Audio(label="Upload Audio File", type="filepath", source="upload"),
|
| 59 |
-
gr.components.Textbox(label="m3u8 URL | E.g.: from kvf.fo or logting.fo")
|
| 60 |
-
],
|
| 61 |
-
outputs=[
|
| 62 |
-
gr.components.Textbox(type="text"),
|
| 63 |
-
"state"
|
| 64 |
-
],
|
| 65 |
|
| 66 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 67 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 68 |
|
| 69 |
demo.launch()
|
|
|
|
| 7 |
# Check if GPU is available
|
| 8 |
use_gpu = torch.cuda.is_available()
|
| 9 |
|
|
|
|
| 10 |
# Configure the pipeline to use the GPU if available
|
| 11 |
if use_gpu:
|
| 12 |
+
p = pipeline("automatic-speech-recognition",
|
| 13 |
+
model="carlosdanielhernandezmena/wav2vec2-large-xlsr-53-faroese-100h", device=0)
|
| 14 |
else:
|
| 15 |
+
p = pipeline("automatic-speech-recognition",
|
| 16 |
+
model="carlosdanielhernandezmena/wav2vec2-large-xlsr-53-faroese-100h")
|
|
|
|
| 17 |
|
| 18 |
def extract_audio_from_m3u8(url):
|
| 19 |
try:
|
|
|
|
| 23 |
except Exception as e:
|
| 24 |
return f"An error occurred: {e}"
|
| 25 |
|
| 26 |
+
def transcribe_function(audio, state, uploaded_audio, m3u8_url):
|
|
|
|
| 27 |
if m3u8_url:
|
| 28 |
audio = extract_audio_from_m3u8(m3u8_url)
|
| 29 |
|
|
|
|
| 31 |
audio = uploaded_audio
|
| 32 |
|
| 33 |
if not audio:
|
| 34 |
+
return {state_var: state, transcription_var: state} # Return a meaningful message
|
| 35 |
|
| 36 |
try:
|
| 37 |
time.sleep(3)
|
| 38 |
text = p(audio, chunk_length_s= 50)["text"]
|
| 39 |
state += text + "\n"
|
| 40 |
+
return {state_var: state, transcription_var: state}
|
| 41 |
except Exception as e:
|
| 42 |
+
return {transcription_var: "An error occurred during transcription.", state_var: state} # Handle other exceptions
|
| 43 |
+
|
| 44 |
+
# ... [most of your code remains unchanged]
|
| 45 |
+
|
| 46 |
+
def reset_output(transcription, state):
|
| 47 |
+
"""Function to reset the state to an empty string."""
|
| 48 |
+
return "", ""
|
| 49 |
|
| 50 |
+
with gr.Blocks() as demo:
|
| 51 |
+
state_var = gr.State("")
|
| 52 |
|
| 53 |
+
with gr.Row():
|
| 54 |
+
with gr.Column():
|
| 55 |
+
microphone = gr.Audio(source="microphone", type="filepath", label="Microphone")
|
| 56 |
+
uploaded_audio = gr.Audio(label="Upload Audio File", type="filepath", source="upload")
|
| 57 |
+
m3u8_url = gr.Textbox(label="m3u8 URL | E.g.: from kvf.fo or logting.fo")
|
| 58 |
|
| 59 |
+
with gr.Column():
|
| 60 |
+
transcription_var = gr.Textbox(type="text", label="Transcription", readonly=True)
|
| 61 |
|
| 62 |
+
with gr.Row():
|
| 63 |
+
transcribe_button = gr.Button("Transcribe")
|
| 64 |
+
reset_button = gr.Button("Reset output")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 65 |
|
| 66 |
+
transcribe_button.click(
|
| 67 |
+
transcribe_function,
|
| 68 |
+
[microphone, state_var, uploaded_audio, m3u8_url],
|
| 69 |
+
[transcription_var, state_var]
|
| 70 |
+
)
|
| 71 |
|
| 72 |
+
reset_button.click(
|
| 73 |
+
reset_output,
|
| 74 |
+
[transcription_var, state_var],
|
| 75 |
+
[transcription_var, state_var]
|
| 76 |
+
)
|
| 77 |
|
| 78 |
demo.launch()
|