sound-design

原文涵盖的主题

• Sound design
• How it fits together
• Writing a cue sheet for a new video
• Visual event → sound
• Levels and space
• Adding a generator
• Verify
• Optional: generated audio

Full Production Studio note: this skill drives the toolkit in whaleyxbt/claude-motion. Clone that repo and run your agent inside it for the scripts/paths below to work; the guidance itself is usable anywhere. Most AI-made motion graphics are silent. Sound is what makes a piece feel produced. The sfx/ engine synthesizes every sound in pure Python stdlib: no samples, no licenses, no installs, same output on every run. It reads the same timeline JSON as the picture, so sound can't drift out of sync. | File | Role | |---|---| | sfx/synth.py | generators: each returns mono samples at 48 kHz | | sfx/mix.py | Mix: place sounds in time, master to a loudness target, write WAV | | sfx/loudness.py | BS.1770 integrated loudness (matches ffmpeg's ebur128) | | sfx/cues/ .py | a cue sheet: one video's sound design | Create sfx/cues/ .py: mix.put(t, samples, gain, pan): t always comes from tl, never a typed number. Offsets from a beat (s1["exit"] + 0.05) are fine. mix.typing(start, end, chars): one humanized keystroke per character. mix.putstereo(t, left, right, gain): for stereo sources like pad(). Mastering is automatic: the engine soft-clips and sets gain to hit TARGETLUFS, raising drive only a

这里只展示节选，请以完整原文和当前要求为准。