Real SFX
A Claude Code skill that dresses a motion video with real recorded sounds, taken from your own library, instead of generated ones. The agent finds the moments that need a sound (the project's event list, a HyperFrames composition, a screen recording or the picture alone), picks and processes the recordings, places them on the exact frame at the level the original sounds had, keeps your music and voice, then returns a new MP4, a separate sound effects track in WAV, a report of the sounds placed and a self-critique scoreboard. It ships with no sound library: only 78 interface sounds under CC0, themselves made by synthesis, are offered as an option. Clipwing repo created in October 2026, 113 stars on November 12, 2026. Scripts under the MIT licence, stated in the skill's README: the repo has no LICENSE file.
Strengths
- Finds the moments that need a sound from four sources: the project's event list, a HyperFrames composition, a screen recording or the picture alone
- Always returns a separate sound effects track in WAV on top of the MP4: you rework the mix in your editing software
- Keeps the original music and voice, and takes the synthesized sound layer out instead of doubling it
- Starts from the fact that the agent cannot hear: a listening page lets you settle in one tap each blind comparison and each moment its critique flags
Limitations
- Ships with no sound library: without yours, you only get 78 interface sounds (clicks, notifications), themselves made by synthesis, and nothing for a hit or a whoosh
- The MIT licence of the scripts is only written in the skill README: no LICENSE file, so GitHub shows the repo as unlicensed
- Manual setup: Python 3.9 or later and FFmpeg required, plus Node.js for the 78 optional sounds, and on macOS the full FFmpeg build, without which two filters are missing
- A repo from October 2026 with 113 stars: few usage reports beyond its author
Best for
- Editors and motion designers who already own a sound effects library and have an agent assemble their videos (HyperFrames, screen recordings)
- Replacing the synthesized sounds of an already rendered video with real recordings, without touching the music or the voice
- Adding sound to a screen recording: keystrokes, clicks and notifications placed on the right moments