PromptClip-Skill: a prompt-driven filter for throwaway family videos
I built PromptClip-Skill, an open-source Codex Skill for finding the meaningful moments inside a folder of casual videos. This started with a very ordinary problem: recording family life is easy, but reviewing it is not. A few minutes of a baby playing, a beach walk, or a weekend outing can produce dozens of clips. Many contain camera shake, accidental recordings, repeated actions, empty frames,…
PromptClip-Skill is an open-source Codex Skill designed to sift through a folder of casual videos and highlight the meaningful moments. The inspiration behind this tool stems from the common issue of recording family life being effortless, but reviewing it can be quite challenging. A simple few minutes of a baby playing, a beach walk, or a weekend outing can generate dozens of clips, but many of these contain video imperfections such as camera shake, accidental recordings, repeated actions, empty frames, or subjects leaving the frame.
The core concept of PromptClip-Skill revolves around filtering the content before dealing with editing. It utilizes a natural language selection rule, such as keeping moments where the baby is clearly visible and reacting naturally, laughing, interacting with a parent, or completing a meaningful action, while discarding shaky footage, repeated scenes, empty frames, and incomplete actions.
The prompt provided to the AI describes both positive signals and exclusions. The output generated by this tool is a concise list of candidate clips along with timestamps and an edit decision list that can be reviewed before final export. This workflow proves beneficial for personal memories where the selection quality and transparency hold more importance than automatic transitions.
The creator built this tool because while one-click video creation tools are convenient, they often lack the necessary control for this specific use case. Custom prompts are not always adhered to precisely, negative requirements such as removing all repeated or empty footage are difficult to express, and the results can frequently follow a generic template rather than the original memory structure.
For a beach walk with a child, the user might want the first discovery of water, a clear facial expression, a complete interaction with a parent, and a few coherent walking shots, without wanting to include every visually active second.
Running locally in Codex, PromptClip-Skill first requires pointing it at a local folder of videos. The user should then describe the people, actions, mood, duration, and exclusions. After analyzing the footage and identifying candidate moments, the user can review the timestamps and the generated edit decision list. The selected clips can then be exported using FFmpeg or further edited in another tool, ensuring the original media remains untouched.
The intermediate results are also accessible, allowing for adjustments to the prompt if needed. Importantly, this tool is not limited to family videos. The same approach can be applied to travel, pets, sports, events, interviews, lessons, or any collection of media where meaningful moments are surrounded by throwaway footage. As a small open-source experiment, PromptClip-Skill is not a hosted video SaaS or a substitute for a professional NLE.
Rather, it aims to make prompt-based selection practical and repeatable. The project repository is available at https://github.com/ron0115/PromptClip-Skill. The creator is particularly interested in feedback on improving highlight scoring, prompt templates, local model support, and enhancing the review/export process for better reliability.
Ultimately, the larger question this tool seeks to address is whether an AI editor can be more beneficial when it initially learns what to discard.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
