Spaces:
Sleeping
Sleeping
File size: 2,887 Bytes
5f2a5b3 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 | import pandas as pd
from asr.speech_2_span import whisperx_to_word_df
from helpers.build_span import build_spans
from llm_pipeline.call_llm import batch_classify_async_llm, span_classify_llm
from helpers.merge_intervals import merge_intervals
from asr.mute_audio import mute_audio
def main():
AUDIO_ID = "audio_001"
AUDIO_PATH = "C:/Users/User/Downloads/hate_videos/hate_videos/hate_video_3.mp4"
EXTRACTED_WAV_PATH = "C:/python_proj/git/AutoCenzurer/artifacts/hate_video_3_extracted.wav"
OUTPUT_PATH = "C:/python_proj/git/AutoCenzurer/artifacts/hate_video_3_output.mp4"
SPANS_CSV = "artifacts/spans.csv"
WORDS_CSV = "artifacts/words.csv"
# 1 Audio / Video - Audio extraction (.wav) - ASR (Whisper) - words_df -
# 2 build_spans() - span_df -
# 3 deterministic signals - min_allowed_label -
# 4 LLM classification - final_enforced_label -
# 5 filter label != NONE - extract intervals -
# 6 merge_intervals() - mute_audio() ###
print("1 STEP: Running ASR")
# words_df = whisperx_to_word_df(
# audio_path=AUDIO_PATH,
# audio_id=AUDIO_ID
# )
# words_df.to_csv(WORDS_CSV, index=False)
# words_df = pd.read_csv(WORDS_CSV)
print("2 STEP: Building spans")
# spans_df = build_spans(words_df)
# spans_df.to_csv("artifacts/spans_async.csv", index=False)
spans_df = pd.read_csv("artifacts/spans_async.csv")
print(f"LOG: Spans saved to {SPANS_CSV}")
print("3+4 STEP: Deterministic Signals Analysis & Classifying spans with LLM API")
### SYNC WORKING STEP###
# for idx, row in spans_df.iterrows():
# result = classify_span_with_llm(
# span_text=row["span_text"],
# has_excessive_profanity=has_excessive_profanity,
# has_slur=has_slur,
# has_targeted_insult=has_targeted_insult
# )
# for k, v in result.items():
# spans_df.at[idx, k] = v
### ASYNC IMPLEMENTATION ###
spans_llm_df = batch_classify_async_llm(spans_df)
spans_llm_df.to_csv("artifacts/spans_async_llm.csv", index=False)
print("5 STEP Filtering and Extracy harmful spans")
harmful_spans = spans_df[
spans_df["final_enforced_label"] != "NONE"
]
if harmful_spans.empty:
print("No harmful spans detected. Nice.")
return
intervals = list(zip(harmful_spans["start_time"], harmful_spans["end_time"]))
print("6 STEP: Merge & MUTE")
merged_intervals = merge_intervals(
intervals,
pad_before=0.5,
pad_after=0.8
)
print("Merged intervals:")
for i in merged_intervals:
print(" ", i)
mute_audio(
audio_path=EXTRACTED_WAV_PATH,
output_path=OUTPUT_PATH,
intervals=merged_intervals
)
print(f"Muted audio written to {OUTPUT_PATH}")
if __name__ == "__main__":
main() |