Spaces:
Build error
Build error
Commit
·
778f982
1
Parent(s):
56811cc
Update app.py
Browse files
app.py
CHANGED
@@ -37,7 +37,7 @@ def segment_to_dict(segment):
|
|
37 |
return segment
|
38 |
|
39 |
def download_video(video_url: str):
|
40 |
-
download_convert_video_to_audio(yt_dlp, video_url, f"
|
41 |
|
42 |
def transcribe_video(video_url: str, word_timestamps: bool = True, model_size: str = "tiny"):
|
43 |
print(word_timestamps)
|
@@ -47,8 +47,8 @@ def transcribe_video(video_url: str, word_timestamps: bool = True, model_size: s
|
|
47 |
print("getting hex")
|
48 |
rand_id = uuid.uuid4().hex
|
49 |
print("doing download")
|
50 |
-
download_convert_video_to_audio(yt_dlp, video_url, f"
|
51 |
-
segments, info = model.transcribe(f"
|
52 |
segments = [segment_to_dict(segment) for segment in segments]
|
53 |
total_duration = round(info.duration, 2) # Same precision as the Whisper timestamps.
|
54 |
with tqdm(total=total_duration, unit=" seconds") as pbar:
|
@@ -58,7 +58,7 @@ def transcribe_video(video_url: str, word_timestamps: bool = True, model_size: s
|
|
58 |
|
59 |
print(pbar)
|
60 |
print(info)
|
61 |
-
os.remove(f"
|
62 |
print("Detected language '%s' with probability %f" % (info.language, info.language_probability))
|
63 |
print(segments)
|
64 |
return segments
|
|
|
37 |
return segment
|
38 |
|
39 |
def download_video(video_url: str):
|
40 |
+
download_convert_video_to_audio(yt_dlp, video_url, f"{uuid.uuid4().hex}")
|
41 |
|
42 |
def transcribe_video(video_url: str, word_timestamps: bool = True, model_size: str = "tiny"):
|
43 |
print(word_timestamps)
|
|
|
47 |
print("getting hex")
|
48 |
rand_id = uuid.uuid4().hex
|
49 |
print("doing download")
|
50 |
+
download_convert_video_to_audio(yt_dlp, video_url, f"{rand_id}")
|
51 |
+
segments, info = model.transcribe(f"{rand_id}.mp3", beam_size=5, word_timestamps=word_timestamps)
|
52 |
segments = [segment_to_dict(segment) for segment in segments]
|
53 |
total_duration = round(info.duration, 2) # Same precision as the Whisper timestamps.
|
54 |
with tqdm(total=total_duration, unit=" seconds") as pbar:
|
|
|
58 |
|
59 |
print(pbar)
|
60 |
print(info)
|
61 |
+
os.remove(f"{rand_id}.mp3")
|
62 |
print("Detected language '%s' with probability %f" % (info.language, info.language_probability))
|
63 |
print(segments)
|
64 |
return segments
|