What problem does it solve?
Audio tracks placed in a HyperFrames video composition often fight each other — a music bed drowns out a voiceover, levels drift, and raw recordings sound amateur. This Skill mixes already-placed audio by writing effect chains, volume automation, and voiceover carve settings directly onto the HTML media elements, so preview and render produce the same processed sound.
Core Features & Use Cases
- Effect chains via data-fx-chain: Apply filters (highpass, peaking, shelves), dynamics (compressor, limiter, gate), nonlinear effects (saturate, bitcrush), and time effects (delay, reverb, chorus, phaser) in signal order on any audio or video element.
- Voiceover carve: Automatically analyze voice tracks and cut only the bands they occupy from the music bed, with dynamic level matching, using the carve.mjs script or Studio panel.
- Automation envelopes: Drive track volume or automatable effect parameters over time with breakpoint lanes in data-automation, including submix buses via <hf-audio-group>.
- Diagnosis and presets: Measure problem audio with ffmpeg-based recipes and apply named presets and jobs (Reduce Mud, Add Clarity, voice-clean) mapped to audible symptoms.
- Use Case: You have a narration track and a background music bed in a composition. Run node scripts/carve.mjs --comp index.html to detect the bed and voices, carve the bed dynamically at strength 0.25, and write the resulting chain and automation back into the HTML.
Quick Start
Ask the agent to carve the music bed under the voiceover in your HyperFrames composition by running the carve script against your index.html file.
