audiocut-keyword

Automated keyword filtering works by transcribing audio into 30-second segments with FunASR, identifying configured keyword occurrences with character-level timestamps, and using FFmpeg clipping to remove those segments from the final audio output.

Yes, a keyword configuration file is required to define target terms. The pipeline uses this file to locate matching occurrences in the transcript and generate a precise deletion plan for FFmpeg-based clipping.

Yes, the filtering mechanism supports adjustable pre and post buffers around detected keywords. This prevents clipping adjacent words during the FFmpeg-based audio segment removal process.

The FunASR transcription pipeline processes audio inputs in 30-second segments to identify keyword occurrences, supporting batch processing of audio assets like interviews and podcasts to generate filtered output files.

Transcription-driven keyword discovery identifies target terms with character-level timestamps for precise editing. This generates a deletion plan applied via FFmpeg clipping to produce clean, moderated video podcast audio.

If pre and post buffers are not adjusted properly, adjacent words may be clipped or missed during editing. Ensure the keyword configuration file is accurate and the 30-second FunASR transcription pipeline captures all occurrences for the deletion plan.