waveform-silence

What problem does it solve?

This Skill helps editors and agents detect silences in audio/video clips when fixed dB threshold methods fail due to inconsistent background noise, room changes, or intermittent ambient sounds, producing reliable trim specifications per clip.

Core Features & Use Cases

• Per-clip visual analysis: Generate chunked waveform images and determine a noise-floor baseline for each clip rather than using a global threshold.
• Robust silence detection: Identify sustained speech vs flat/noise regions, merge short gaps, and avoid cutting transient spikes or mid-word dips.
• Seamless integration: Convert visual boundaries into timestamped keeps and call crop_spec to produce trimmed spec JSON for downstream steps like rm_fillers or concat.
• Use Case: Recover clean speech regions from multi-clip recordings where fans, music, or mic placement changes make automated threshold trimming unreliable.

Quick Start

Use waveform-silence to generate waveform images for each clip, visually identify silences relative to each clip's noise floor, and output crop_spec trim specs for cropping.