Spaces:
Paused
Paused
WIP
Browse files
app.py
CHANGED
@@ -248,22 +248,17 @@ def transcribe_youtube(url, return_timestamps, generate_subs, chunk_length_s=30,
|
|
248 |
logger.error(f"Invalid audio URL scheme from Sieve: {audio_url}")
|
249 |
raise gr.Error("Invalid audio URL scheme received from Sieve")
|
250 |
|
251 |
-
# Prepare request parameters
|
252 |
-
params = {
|
253 |
-
"source_language": "tg", # Tajik language
|
254 |
-
"timestamp_level": "sentence" if return_timestamps else None,
|
255 |
-
"task": "transcribe",
|
256 |
-
"chunk_length_s": chunk_length_s,
|
257 |
-
"batch_size": batch_size
|
258 |
-
}
|
259 |
-
logger.info(f"Prepared API request parameters: {params}")
|
260 |
-
|
261 |
# Send request to API
|
262 |
logger.info("Sending transcription request to API...")
|
263 |
response = requests.post(
|
264 |
f"{API_URL}/transcribe/url",
|
265 |
-
json={
|
266 |
-
|
|
|
|
|
|
|
|
|
|
|
267 |
timeout=1800
|
268 |
)
|
269 |
response.raise_for_status()
|
@@ -324,22 +319,17 @@ def transcribe(inputs, return_timestamps, generate_subs, chunk_length_s=30, batc
|
|
324 |
with open(inputs, "rb") as f:
|
325 |
files = {"file": f}
|
326 |
|
327 |
-
# Prepare request parameters
|
328 |
-
params = {
|
329 |
-
"source_language": "tg", # Tajik language
|
330 |
-
"timestamp_level": "sentence" if return_timestamps else None,
|
331 |
-
"task": "transcribe",
|
332 |
-
"chunk_length_s": chunk_length_s,
|
333 |
-
"batch_size": batch_size
|
334 |
-
}
|
335 |
-
logger.info(f"Prepared API request parameters: {params}")
|
336 |
-
|
337 |
# Send request to API
|
338 |
logger.info("Sending transcription request to API...")
|
339 |
response = requests.post(
|
340 |
f"{API_URL}/transcribe",
|
341 |
files=files,
|
342 |
-
|
|
|
|
|
|
|
|
|
|
|
343 |
timeout=1800
|
344 |
)
|
345 |
response.raise_for_status()
|
@@ -418,7 +408,7 @@ mf_transcribe = gr.Interface(
|
|
418 |
gr.Audio(sources="microphone", type="filepath"),
|
419 |
gr.Checkbox(label="Include timestamps", value=True),
|
420 |
gr.Checkbox(label="Generate subtitles", value=True),
|
421 |
-
gr.Slider(minimum=10, maximum=
|
422 |
gr.Slider(minimum=32, maximum=256, value=128, step=32, label="Batch Size")
|
423 |
],
|
424 |
outputs=[
|
@@ -439,7 +429,7 @@ file_transcribe = gr.Interface(
|
|
439 |
gr.Audio(sources="upload", type="filepath", label="Audio file"),
|
440 |
gr.Checkbox(label="Include timestamps", value=True),
|
441 |
gr.Checkbox(label="Generate subtitles", value=True),
|
442 |
-
gr.Slider(minimum=10, maximum=
|
443 |
gr.Slider(minimum=32, maximum=256, value=8, step=32, label="Batch Size")
|
444 |
],
|
445 |
outputs=[
|
|
|
248 |
logger.error(f"Invalid audio URL scheme from Sieve: {audio_url}")
|
249 |
raise gr.Error("Invalid audio URL scheme received from Sieve")
|
250 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
251 |
# Send request to API
|
252 |
logger.info("Sending transcription request to API...")
|
253 |
response = requests.post(
|
254 |
f"{API_URL}/transcribe/url",
|
255 |
+
json={
|
256 |
+
"url": audio_url,
|
257 |
+
"timestamp_level": "sentence" if return_timestamps else None,
|
258 |
+
"task": "transcribe",
|
259 |
+
"chunk_length_s": chunk_length_s,
|
260 |
+
"batch_size": batch_size
|
261 |
+
},
|
262 |
timeout=1800
|
263 |
)
|
264 |
response.raise_for_status()
|
|
|
319 |
with open(inputs, "rb") as f:
|
320 |
files = {"file": f}
|
321 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
322 |
# Send request to API
|
323 |
logger.info("Sending transcription request to API...")
|
324 |
response = requests.post(
|
325 |
f"{API_URL}/transcribe",
|
326 |
files=files,
|
327 |
+
data={
|
328 |
+
"timestamp_level": "sentence" if return_timestamps else None,
|
329 |
+
"task": "transcribe",
|
330 |
+
"chunk_length_s": chunk_length_s, # Send as integer
|
331 |
+
"batch_size": batch_size # Send as integer
|
332 |
+
},
|
333 |
timeout=1800
|
334 |
)
|
335 |
response.raise_for_status()
|
|
|
408 |
gr.Audio(sources="microphone", type="filepath"),
|
409 |
gr.Checkbox(label="Include timestamps", value=True),
|
410 |
gr.Checkbox(label="Generate subtitles", value=True),
|
411 |
+
gr.Slider(minimum=10, maximum=30, value=30, step=5, label="Chunk Length (seconds)"),
|
412 |
gr.Slider(minimum=32, maximum=256, value=128, step=32, label="Batch Size")
|
413 |
],
|
414 |
outputs=[
|
|
|
429 |
gr.Audio(sources="upload", type="filepath", label="Audio file"),
|
430 |
gr.Checkbox(label="Include timestamps", value=True),
|
431 |
gr.Checkbox(label="Generate subtitles", value=True),
|
432 |
+
gr.Slider(minimum=10, maximum=30, value=15, step=5, label="Chunk Length (seconds)"),
|
433 |
gr.Slider(minimum=32, maximum=256, value=8, step=32, label="Batch Size")
|
434 |
],
|
435 |
outputs=[
|