Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -31,6 +31,9 @@ CHUNK_LENGTH_S = 30 # Split long audio into 30-second chunks
|
|
| 31 |
SAMPLE_RATE = 16000 # Whisper expects 16kHz audio
|
| 32 |
YT_LENGTH_LIMIT_S = 3600 # Limit YouTube videos to 1 hour
|
| 33 |
|
|
|
|
|
|
|
|
|
|
| 34 |
# Lazy load state for the Whisper model
|
| 35 |
_WHISPER_STATE = {"initialized": False, "pipe": None, "device": "cpu"}
|
| 36 |
|
|
@@ -451,50 +454,51 @@ with gr.Blocks(title="Whisper-ASR") as demo:
|
|
| 451 |
api_name="transcribe",
|
| 452 |
)
|
| 453 |
|
| 454 |
-
# Tab 2: YouTube
|
| 455 |
-
|
| 456 |
-
with gr.
|
| 457 |
-
with gr.
|
| 458 |
-
|
| 459 |
-
|
| 460 |
-
|
| 461 |
-
|
| 462 |
-
|
| 463 |
-
|
| 464 |
-
with gr.Row():
|
| 465 |
-
yt_task_radio = gr.Radio(
|
| 466 |
-
choices=["transcribe", "translate"],
|
| 467 |
-
value="transcribe",
|
| 468 |
-
label="Task",
|
| 469 |
-
info="Translate converts any language to English",
|
| 470 |
-
)
|
| 471 |
-
yt_language_dropdown = gr.Dropdown(
|
| 472 |
-
choices=LANGUAGES,
|
| 473 |
-
value="auto",
|
| 474 |
-
label="Language",
|
| 475 |
)
|
| 476 |
|
| 477 |
-
|
| 478 |
-
|
| 479 |
-
|
| 480 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 481 |
|
| 482 |
-
|
| 483 |
|
| 484 |
-
|
| 485 |
-
|
| 486 |
-
|
| 487 |
-
|
| 488 |
-
|
| 489 |
-
|
| 490 |
-
|
| 491 |
|
| 492 |
-
|
| 493 |
-
|
| 494 |
-
|
| 495 |
-
|
| 496 |
-
|
| 497 |
-
|
| 498 |
|
| 499 |
|
| 500 |
if __name__ == "__main__":
|
|
|
|
| 31 |
SAMPLE_RATE = 16000 # Whisper expects 16kHz audio
|
| 32 |
YT_LENGTH_LIMIT_S = 3600 # Limit YouTube videos to 1 hour
|
| 33 |
|
| 34 |
+
# Detect if running on Hugging Face Spaces (YouTube won't work there due to network restrictions)
|
| 35 |
+
IS_HF_SPACE = os.environ.get("SPACE_ID") is not None
|
| 36 |
+
|
| 37 |
# Lazy load state for the Whisper model
|
| 38 |
_WHISPER_STATE = {"initialized": False, "pipe": None, "device": "cpu"}
|
| 39 |
|
|
|
|
| 454 |
api_name="transcribe",
|
| 455 |
)
|
| 456 |
|
| 457 |
+
# Tab 2: YouTube (only shown when running locally)
|
| 458 |
+
if not IS_HF_SPACE:
|
| 459 |
+
with gr.TabItem("YouTube"):
|
| 460 |
+
with gr.Row():
|
| 461 |
+
with gr.Column():
|
| 462 |
+
yt_url_input = gr.Textbox(
|
| 463 |
+
label="YouTube URL",
|
| 464 |
+
placeholder="Paste a YouTube video URL here...",
|
| 465 |
+
lines=1,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 466 |
)
|
| 467 |
|
| 468 |
+
with gr.Row():
|
| 469 |
+
yt_task_radio = gr.Radio(
|
| 470 |
+
choices=["transcribe", "translate"],
|
| 471 |
+
value="transcribe",
|
| 472 |
+
label="Task",
|
| 473 |
+
info="Translate converts any language to English",
|
| 474 |
+
)
|
| 475 |
+
yt_language_dropdown = gr.Dropdown(
|
| 476 |
+
choices=LANGUAGES,
|
| 477 |
+
value="auto",
|
| 478 |
+
label="Language",
|
| 479 |
+
)
|
| 480 |
+
|
| 481 |
+
yt_timestamps_checkbox = gr.Checkbox(
|
| 482 |
+
label="Return Timestamps",
|
| 483 |
+
value=False,
|
| 484 |
+
)
|
| 485 |
|
| 486 |
+
yt_transcribe_btn = gr.Button("Transcribe YouTube", variant="primary")
|
| 487 |
|
| 488 |
+
with gr.Column():
|
| 489 |
+
yt_embed = gr.HTML(label="Video")
|
| 490 |
+
yt_output = gr.Textbox(
|
| 491 |
+
label="Transcription",
|
| 492 |
+
placeholder="Transcribed text will appear here...",
|
| 493 |
+
lines=10,
|
| 494 |
+
)
|
| 495 |
|
| 496 |
+
yt_transcribe_btn.click(
|
| 497 |
+
fn=transcribe_youtube_streaming,
|
| 498 |
+
inputs=[yt_url_input, yt_task_radio, yt_language_dropdown, yt_timestamps_checkbox],
|
| 499 |
+
outputs=[yt_embed, yt_output],
|
| 500 |
+
api_name="transcribe_youtube",
|
| 501 |
+
)
|
| 502 |
|
| 503 |
|
| 504 |
if __name__ == "__main__":
|