[{"data":1,"prerenderedAt":4},["ShallowReactive",2],{"readme:video-spec-builder":3},"\u003Cimg width=\"2172\" height=\"724\" alt=\"ChatGPT Image May 16, 2026, 10_46_58 PM\" src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F7820d93e-84b6-4e09-904c-9567c6595c57\" \u002F>\u003Cp>\u003Cstrong>English\u003C\u002Fstrong> · \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Ffeicaiclub\u002Fvideo-spec-builder\u002Fblob\u002FHEAD\u002FREADME.zh.md\" rel=\"nofollow ugc noopener\">中文\u003C\u002Fa>\u003C\u002Fp>\n\u003Ch1>video-spec-builder\u003C\u002Fh1>\n\u003Cp>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Ffeicaiclub\u002Fvideo-spec-builder\u002Fblob\u002FHEAD\u002FLICENSE\" rel=\"nofollow ugc noopener\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-MIT-brightgreen\" alt=\"License: MIT\" \u002F>\u003C\u002Fa> \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FAgent-Agnostic-blueviolet\" alt=\"Agent Agnostic\" \u002F> \u003Ca href=\"https:\u002F\u002Fskills.sh\" rel=\"nofollow ugc noopener\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fskills.sh-Compatible-brightgreen\" alt=\"skills.sh Compatible\" \u002F>\u003C\u002Fa>\u003C\u002Fp>\n\u003Cblockquote>\n\u003Cp>A skill that works like a video director. You say \"I want to make a video,\" and it grills you with questions until your idea is a script you can actually shoot.\u003C\u002Fp>\n\u003C\u002Fblockquote>\n\u003Cp>I built this skill after realizing the hard part of making a video isn't the rendering. It's figuring out what you actually want.\u003C\u002Fp>\n\u003Cp>You've got a vague idea in your head: a product video, a short for social, a company intro. But it's fuzzy. The moment you try to build it, the details get you — how long each shot runs, what's on screen, what comes first and what comes later. You probably haven't pinned them all down, and you might not even be able to put them into words.\u003C\u002Fp>\n\u003Cp>video-spec-builder gets you through that part. Install it, then tell your AI \"I want to make a video\" inside Codex or Claude Code, and it takes over the conversation. It listens to your brief the way a director would, then keeps asking: Who's this for? How long? What's the one line people should walk away with? Which shot carries the weight? Anywhere you go vague, or skip something, it stops and pushes you to fill it in.\u003C\u002Fp>\n\u003Cp>A few rounds of that, and the fuzzy idea becomes a \u003Ccode>video-spec.md\u003C\u002Fcode>: a shot-by-shot script, timed to the second, every shot written out. Hand that to HyperFrames and it renders into a real video.\u003C\u002Fp>\n\u003Cp>It won't shoot the video for you, and it won't invent the idea. It does one thing: push you, and stay with you, until the idea is something you can actually build.\u003C\u002Fp>\n\u003Ch2>What it helps with\u003C\u002Fh2>\n\u003Cp>The problem it solves is \"I have an idea but I can't explain it.\" A few situations where it earns its keep:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>You know the feeling you want but can't describe the actual picture. It refuses words like \"premium\" or \"high-impact\" and keeps after you until you can describe real shots and real motion.\u003C\u002Fli>\n\u003Cli>You have an idea but never thought parts of it through. Maybe you've got the opening and the ending but not the middle. Maybe it never crossed your mind that a section could use captions, or that visuals can move to the beat of the music. It brings those up.\u003C\u002Fli>\n\u003Cli>You have plenty of raw material but no order to it. A script, selling points, a pile of assets — it helps you cut that into individual shots and put them in sequence.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>In the end it writes all of it into a script: what each shot shows, how it's presented, how long it holds, how it cuts to the next one.\u003C\u002Fp>\n\u003Cp>There are two ways to use it. With no script yet, it talks you through the whole thing from scratch and produces a \u003Ccode>video-spec.md\u003C\u002Fcode>. With a script already there and just one thing to change, you tell it what you want different; it asks enough to be sure, makes the change, and checks whether it knocked anything else loose.\u003C\u002Fp>\n\u003Ch2>The workflow\u003C\u002Fh2>\n\u003Cp>It's two skills working in sequence. video-spec-builder sits upstream and turns your idea into a script. HyperFrames sits downstream and turns the script into video.\u003C\u002Fp>\n\u003Cpre>\u003Ccode>       You: \"I want to make a video\"\n                │\n                ▼\n   ┌────────────────────────┐\n   │   video-spec-builder   │   asks, breaks it into shots\n   └────────────────────────┘\n                │\n                ▼\n          video-spec.md           shot-by-shot script, timed\n                │\n                ▼   \u002Fhyperframes\n   ┌────────────────────────┐\n   │       HyperFrames      │   renders from the script\n   └────────────────────────┘\n                │\n                ▼\n          finished video\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>So before you start, you'll want both skills installed.\u003C\u002Fp>\n\u003Ch2>Install\u003C\u002Fh2>\n\u003Cp>I mostly use this skill in \u003Cstrong>Codex\u003C\u002Fstrong>, and after that \u003Cstrong>Claude Code\u003C\u002Fstrong>. Those are the two setups\u003C\u002Fp>\n",1784240408339]