Update utils.py
Browse files
utils.py
CHANGED
|
@@ -7,25 +7,27 @@ import subprocess
|
|
| 7 |
# Load Whisper model
|
| 8 |
model = whisper.load_model("base")
|
| 9 |
|
| 10 |
-
def process_video(video_path, language
|
| 11 |
output_video_path = os.path.join(tempfile.gettempdir(), "converted_video.mp4")
|
| 12 |
srt_path = os.path.join(tempfile.gettempdir(), "subtitles.srt")
|
| 13 |
|
| 14 |
try:
|
| 15 |
# Convert video to MP4 using ffmpeg
|
| 16 |
-
|
|
|
|
| 17 |
subprocess.run(
|
| 18 |
["ffmpeg", "-i", video_path, "-c:v", "libx264", "-preset", "fast", output_video_path],
|
| 19 |
-
check=True,
|
| 20 |
stdout=subprocess.PIPE,
|
| 21 |
stderr=subprocess.PIPE
|
| 22 |
)
|
| 23 |
-
print("Video converted successfully!")
|
| 24 |
|
| 25 |
# Transcribe video
|
| 26 |
-
|
|
|
|
| 27 |
result = model.transcribe(output_video_path, language="en")
|
| 28 |
-
|
|
|
|
| 29 |
|
| 30 |
# Translation logic
|
| 31 |
segments = []
|
|
@@ -46,9 +48,9 @@ def process_video(video_path, language): # Accept file path, not file object
|
|
| 46 |
}
|
| 47 |
model_name = model_map.get(language)
|
| 48 |
if not model_name:
|
| 49 |
-
return
|
| 50 |
|
| 51 |
-
|
| 52 |
if language == "Telugu":
|
| 53 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
| 54 |
translation_model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
|
|
@@ -70,13 +72,16 @@ def process_video(video_path, language): # Accept file path, not file object
|
|
| 70 |
segments.append({"text": translated_text, "start": segment["start"], "end": segment["end"]})
|
| 71 |
|
| 72 |
# Create SRT file
|
|
|
|
|
|
|
| 73 |
with open(srt_path, "w", encoding="utf-8") as f:
|
| 74 |
for i, segment in enumerate(segments, 1):
|
| 75 |
start = f"{segment['start']:.3f}".replace(".", ",")
|
| 76 |
end = f"{segment['end']:.3f}".replace(".", ",")
|
| 77 |
text = segment["text"].strip()
|
| 78 |
f.write(f"{i}\n00:00:{start} --> 00:00:{end}\n{text}\n\n")
|
| 79 |
-
|
|
|
|
| 80 |
return srt_path
|
| 81 |
|
| 82 |
except subprocess.CalledProcessError as e:
|
|
|
|
| 7 |
# Load Whisper model
|
| 8 |
model = whisper.load_model("base")
|
| 9 |
|
| 10 |
+
def process_video(video_path, language, progress=None):
|
| 11 |
output_video_path = os.path.join(tempfile.gettempdir(), "converted_video.mp4")
|
| 12 |
srt_path = os.path.join(tempfile.gettempdir(), "subtitles.srt")
|
| 13 |
|
| 14 |
try:
|
| 15 |
# Convert video to MP4 using ffmpeg
|
| 16 |
+
if progress:
|
| 17 |
+
progress(0.2, desc="π Converting video to MP4...")
|
| 18 |
subprocess.run(
|
| 19 |
["ffmpeg", "-i", video_path, "-c:v", "libx264", "-preset", "fast", output_video_path],
|
| 20 |
+
check=True,
|
| 21 |
stdout=subprocess.PIPE,
|
| 22 |
stderr=subprocess.PIPE
|
| 23 |
)
|
|
|
|
| 24 |
|
| 25 |
# Transcribe video
|
| 26 |
+
if progress:
|
| 27 |
+
progress(0.4, desc="π Transcribing audio...")
|
| 28 |
result = model.transcribe(output_video_path, language="en")
|
| 29 |
+
if progress:
|
| 30 |
+
progress(0.6, desc="π Translating subtitles...")
|
| 31 |
|
| 32 |
# Translation logic
|
| 33 |
segments = []
|
|
|
|
| 48 |
}
|
| 49 |
model_name = model_map.get(language)
|
| 50 |
if not model_name:
|
| 51 |
+
return None
|
| 52 |
|
| 53 |
+
# Load translation model
|
| 54 |
if language == "Telugu":
|
| 55 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
| 56 |
translation_model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
|
|
|
|
| 72 |
segments.append({"text": translated_text, "start": segment["start"], "end": segment["end"]})
|
| 73 |
|
| 74 |
# Create SRT file
|
| 75 |
+
if progress:
|
| 76 |
+
progress(0.8, desc="π Generating SRT file...")
|
| 77 |
with open(srt_path, "w", encoding="utf-8") as f:
|
| 78 |
for i, segment in enumerate(segments, 1):
|
| 79 |
start = f"{segment['start']:.3f}".replace(".", ",")
|
| 80 |
end = f"{segment['end']:.3f}".replace(".", ",")
|
| 81 |
text = segment["text"].strip()
|
| 82 |
f.write(f"{i}\n00:00:{start} --> 00:00:{end}\n{text}\n\n")
|
| 83 |
+
if progress:
|
| 84 |
+
progress(1.0, desc="β
Done!")
|
| 85 |
return srt_path
|
| 86 |
|
| 87 |
except subprocess.CalledProcessError as e:
|