[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-83156":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":9,"language":10,"languages":9,"totalLinesOfCode":9,"stars":11,"forks":12,"watchers":13,"openIssues":14,"contributorsCount":14,"subscribersCount":14,"size":14,"stars1d":15,"stars7d":16,"stars30d":16,"stars90d":14,"forks30d":14,"starsTrendScore":17,"compositeScore":18,"rankGlobal":9,"rankLanguage":9,"license":19,"archived":20,"fork":20,"defaultBranch":21,"hasWiki":22,"hasPages":20,"topics":23,"createdAt":9,"pushedAt":9,"updatedAt":39,"readmeContent":40,"aiSummary":41,"trendingCount":14,"starSnapshotCount":14,"syncStatus":42,"lastSyncTime":43,"discoverSource":44},83156,"ai-shortfilm-prompts","jnMetaCode\u002Fai-shortfilm-prompts","jnMetaCode","Methodology + prompts + Claude Code Skill behind Zombie Scavenger by Mx-Shell — the AI short PJ Ace called \"one of the best short films I've seen in years.\" Works with Sora · Kling · Veo · Seedance.",null,"Python",140,25,1,0,3,42,20,4.24,"MIT License",false,"main",true,[24,25,26,27,28,29,30,31,32,33,34,35,36,37,38],"ai-video","chinese","claude-code","claude-skills","jimeng","kling","mx-shell","prompt-engineering","seedance","shortfilm","sora","text-to-video","veo","video-prompts","zombie-scavenger","2026-06-12 02:04:31","# ai-shortfilm-prompts\n\n\u003C!-- ════════════════════════════════════════════════════════════════════\n     README HERO DEMO — drop in the 15s clip here.\n     The clip is generated from .\u002Fassets\u002Fdemo-prompt.md (written by the Skill).\n\n     To enable, produce the clip in Seedance, then EITHER:\n\n     (A) mp4 — drag demo.mp4 into any GitHub issue\u002FPR comment, copy the\n         generated https:\u002F\u002Fuser-images.githubusercontent.com\u002F... URL,\n         then uncomment and paste it here:\n     \u003Cp align=\"center\">\n       \u003Cvideo src=\"PASTE_GITHUB_VIDEO_URL_HERE\" width=\"720\" autoplay loop muted playsinline>\u003C\u002Fvideo>\n       \u003Cbr>\u003Csub>▶ 15s demo — generated from a prompt this repo's Skill wrote (\u003Ca href=\".\u002Fassets\u002Fdemo-prompt.md\">prompt\u003C\u002Fa>)\u003C\u002Fsub>\n     \u003C\u002Fp>\n\n     (B) gif — commit assets\u002Fdemo.gif, then uncomment:\n     \u003Cp align=\"center\">\n       \u003Cimg src=\".\u002Fassets\u002Fdemo.gif\" alt=\"15s demo generated with the 5-stage method\" width=\"720\">\n       \u003Cbr>\u003Csub>▶ 15s demo — generated from a prompt this repo's Skill wrote (\u003Ca href=\".\u002Fassets\u002Fdemo-prompt.md\">prompt\u003C\u002Fa>)\u003C\u002Fsub>\n     \u003C\u002Fp>\n════════════════════════════════════════════════════════════════════ -->\n\n[![License: MIT](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-MIT-yellow.svg)](.\u002FLICENSE)\n[![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FjnMetaCode\u002Fai-shortfilm-prompts?style=social)](https:\u002F\u002Fgithub.com\u002FjnMetaCode\u002Fai-shortfilm-prompts)\n[![X Post](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fread_the_thread-@aibuzhiyu-1DA1F2?logo=x)](https:\u002F\u002Fx.com\u002Faibuzhiyu\u002Fstatus\u002F2056426660577288645)\n[![Plugin](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FClaude_Code-plugin_install-blueviolet)](#install-claude-code)\n\n> The complete methodology + prompt library + Claude Code Skill behind\n> **[*Zombie Scavenger*](https:\u002F\u002Fx.com\u002Faibuzhiyu\u002Fstatus\u002F2056426660577288645)**\n> by Mx-Shell — the AI short that Hollywood director PJ Ace called\n> *\"one of the best short films I've seen in years.\"*\n\n**[中文版 →](.\u002FREADME.zh.md)**\n\n> ⚡ **Quick start:** [one-page cheat sheet](.\u002Fcheatsheet.md) ·\n> [failure→fix gallery](.\u002Fcases.md)\n\n---\n\n## 🎬 The tweet that started it all\n\n> *\"This is one of the best short films I've seen in years.*\n> *Very soon, we'll stop calling it 'AI film' and just call it film.\"*\n>\n> *Film name: Zombie Scavenger by MX-Shell.*\n>\n> — **PJ Ace** ([@PJaccetturo](https:\u002F\u002Fx.com\u002FPJaccetturo)),\n> [May 10, 2026](https:\u002F\u002Fx.com\u002Faibuzhiyu\u002Fstatus\u002F2056426660577288645)\n\n| 13.4M views | 82K likes | 7.4K reposts | 39K bookmarks | 2.3K replies |\n|---|---|---|---|---|\n\n\u003Csub>Stats are from PJ Ace's original tweet ([@PJaccetturo](https:\u002F\u002Fx.com\u002FPJaccetturo), May 10, 2026), as of mid-May 2026.\u003C\u002Fsub>\n\nThis repository is **the complete workflow behind that film**, made\navailable because Mx-Shell himself published his prompt collection\ndocuments and walked through his entire method on a public Douyin\nlivestream.\n\n---\n\n## ⚡ The single-line magic prompt (try it tonight)\n\nCopy this into Sora \u002F Seedance \u002F Kling \u002F Veo. Replace `{{...}}`:\n\n```\nAnamorphic widescreen cinematic. Simulated IMAX film camera +\nPanavision C-series lens (35mm focal, f\u002F4 aperture). Handheld\nshot — extremely subtle, breath-like camera float throughout.\n{{your scene description}}.\nNo score. Production audio only.\n```\n\n**Why this works**: real camera bodies + \"breath-like float\" anchor\nthe AI to actual film aesthetics — not the vague \"cinematic feel\"\nkeyword everyone else uses. Full breakdown in\n[methodology.md](.\u002Fmethodology.md).\n\n---\n\n## ❌ vs ✅ — what the method actually changes\n\nSame idea: *a female mech warrior raises an energy shield in a thunderstorm.*\n\n**❌ The naive prompt** (what most people write):\n\n```\nEpic cinematic shot of a beautiful female mech warrior activating a\nstunning energy shield in the rain. Highly detailed, 4K, photorealistic,\nmovie-quality, dramatic lighting.\n```\n\nVague praise — *epic \u002F stunning \u002F 4K \u002F movie-quality* — gives the model\nnothing concrete to anchor to. You get generic game-CG output.\n\n**✅ With the 5-stage method** (excerpt):\n\n```\nCore theme: gritty hard sci-fi mech | rainy dock | battle-damage aesthetic | energy shield | post-apocalyptic live-action\nAtmosphere: simulated IMAX film camera + Panavision C-series (35mm, f\u002F4). Low-saturation teal base, film grain.\nCamera: handheld — extremely subtle, breath-like float throughout.\n9–12s: hexagonal energy cells light up unevenly, some flicker as if faulty; rain bends around a 2m dome.\nEnding: no dialogue, no light burst — just rain vaporizing on the shield, a lightning flash across the dock.\n```\n\nReal camera\u002Flens names + physical reactions + battle damage + an empty\nending = visceral realism. Full sample with the 10-item self-check:\n[examples\u002F02-skill-output-sample.md](.\u002Fskills\u002Fshortfilm-prompt\u002Fexamples\u002F02-skill-output-sample.md).\n\n---\n\n## The story\n\nMay 2026. A 29-year-old vocational-school graduate from rural Yunnan,\nChina — handle **Mx-Shell** — used **10 days** and ~20,000 RMB of\ncloud credits to make a 3-minute AI short called *Zombie Scavenger*:\nan atomic-punk robot wanders into a beachfront villa after a zombie\napocalypse, meets a confused ostrich, and starts dancing\n1980s-style breakdance moves while kicking a zombie's head across\nthe floor.\n\n> On the cost: the widely-quoted \"3,000 RMB\" figure came from Mx-Shell\n> himself, but on livestream he later revised it to \"tens of thousands \u002F\n> 20k+.\" The real spend was likely well above 3,000 — still far below a\n> comparable live-action shoot.\n\nHollywood director **PJ Ace (@PJaccetturo)** retweeted the film,\ncalling it *\"one of the best short films I've seen in years\"* and started a\nsearch for the author.\n\nA few weeks later Mx-Shell went on a Douyin livestream and **gave\naway his entire workflow** — the prompts, the camera language, the\nfailure modes, the reroll counts.\n\nThis repo is the result of digesting 130,000 characters of his\nmaterials into a structured, reusable system.\n\n---\n\n## What's in here\n\n```\nai-shortfilm-prompts\u002F\n├── README.md              You're here. English entry point.\n├── README.zh.md           Chinese version.\n├── cheatsheet.md          One-page cheat sheet (the whole method at a glance).\n├── cheatsheet.zh.md       Chinese version.\n├── cases.md               Failure→fix gallery: common bad output and the fix.\n├── cases.zh.md            Chinese version.\n├── methodology.md         The 5-stage prompt structure, explained.\n├── methodology.zh.md      Chinese version.\n├── faq.md                 Q&A: tools, failures, costs, edge cases.\n├── faq.zh.md              Chinese version.\n├── credits.md             Sources & attribution.\n├── credits.zh.md          Chinese version.\n├── showcase.md            Things people made with the method.\n├── CONTRIBUTING.md        Submission template & rules for new prompts.\n├── LICENSE                MIT (this work)\n├── NOTICE                 Attribution + Mx-Shell ARR details (dual-licensing)\n│\n├── prompts\u002F                Mx-Shell's complete original prompts.\n│                           Body kept in Chinese (his authorial\n│                           voice), with English header on each file.\n│   ├── README.md           Index of all prompt archives\n│   ├── index.json          Machine-readable index of every prompt\n│   ├── zombie-scavenger.md             *Zombie Scavenger*\n│   ├── kamen-rider-transformations.md   Kamen Rider transformation × 6 (5 riders + flight)\n│   ├── kaisa-transformation.md       LoL Kai'Sa transformation × 3 versions\n│   ├── pacific-rim-gundam.md         Pacific Rim + Gundam mech-drop\n│   ├── cyber-wuxia.md                Shaw Brothers + steampunk wuxia template\n│   └── metal-gear-charge-combat.md   Weapon-charge + combat composite\n│\n├── templates\u002F              IP-stripped generalized templates (EN + .zh.md siblings).\n│                           Authored by jnMetaCode based on Mx-Shell's structure.\n│   ├── 15s-transformation.md         15-second transformation\n│   ├── multi-shot-narrative.md       Multi-shot edited narrative\n│   ├── atmosphere-prefabs.md         8 reusable atmosphere\u002Flook prefabs\n│   └── negative-prompts.md           Reusable negative-prompt prefab (per-model)\n│\n├── assets\u002F                 Diagrams + the README hero-demo prompt.\n│   ├── demo-prompt.md       Copy-paste 15s prompt the Skill wrote (hero slot)\n│   └── 5-stage-structure.svg  The structure diagram\n│\n├── skills\u002Fshortfilm-prompt\u002F   Claude Code Skill\n│   ├── SKILL.md            How Claude should generate prompts (7 hard rules + 10-item checklist)\n│   ├── TESTING.md          How to run rigorous skill tests in another Claude window\n│   └── examples\u002F           4 test cases (5 files) with expected outputs\n│\n└── .claude-plugin\u002F         Plugin metadata (plugin.json + marketplace.json)\n```\n\n---\n\n## TL;DR — The 5-stage prompt structure\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\".\u002Fassets\u002F5-stage-structure.svg\" alt=\"The 5-stage prompt structure\" width=\"660\">\n\u003C\u002Fp>\n\nEvery Mx-Shell video prompt follows the same skeleton. The order matters:\n\n```\n1. Core theme            ← 3-6 style tags separated by |\n2. Character & scene     ← Face \u002F clothing \u002F environment\n3. Atmosphere & quality  ← Visual base \u002F color tone \u002F style core\n4. Camera rules          ← Single-shot or multi-shot \u002F angle \u002F breathing\n5. Storyboard            ← Per-second OR per-shot breakdown\n```\n\n### Three counter-intuitive rules\n\n1. **Specify real camera + lens models.**\n   Don't write *\"cinematic feel\"*. Write\n   *\"simulated IMAX film camera, Panavision C-series lens, 35mm focal,\n   f\u002F4 aperture.\"* AI training data binds those exact strings to real\n   movie aesthetics.\n\n2. **Describe imperfections.**\n   *\"Battle-damaged armor, paint worn off, oil in the joints, minor\n   facial blemishes preserved.\"* Perfection looks fake. The visceral\n   realism comes from the flaws.\n\n3. **Leave the ending empty.**\n   *\"No dialogue. No explosion. No blinding light.\n   Just a figure standing in the smoke, a meteor crossing the sky.\"*\n\nFull methodology in [methodology.md](.\u002Fmethodology.md).\n\n---\n\n## Install (Claude Code)\n\n### Option 1 — Plugin Marketplace ⭐ (one-line install)\n\n```\n\u002Fplugin marketplace add jnMetaCode\u002Fai-shortfilm-prompts\n\u002Fplugin install ai-shortfilm-prompts@ai-shortfilm-prompts\n```\n\nThen in Claude Code:\n\n```\n\u002Fai-shortfilm-prompts:shortfilm-prompt  Help me write a 15-second prompt for a\n                                        robot transformation, green color palette,\n                                        energy core in the belt buckle,\n                                        post-apocalyptic jungle background\n```\n\n### Option 2 — Try it inside this repo\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FjnMetaCode\u002Fai-shortfilm-prompts.git\ncd ai-shortfilm-prompts\nclaude --plugin-dir .   # then: \u002Fai-shortfilm-prompts:shortfilm-prompt\n```\n\n### Option 3 — Make it available globally (manual copy)\n\n```bash\nmkdir -p ~\u002F.claude\u002Fskills\ncp -r ai-shortfilm-prompts\u002Fskills\u002Fshortfilm-prompt \\\n      ~\u002F.claude\u002Fskills\u002F\n```\n\n### Option 4 — As a submodule in your own project\n\n```bash\ngit submodule add https:\u002F\u002Fgithub.com\u002FjnMetaCode\u002Fai-shortfilm-prompts.git \\\n                  vendor\u002Fai-shortfilm-prompts\nclaude --plugin-dir vendor\u002Fai-shortfilm-prompts   # loads the plugin + skill\n```\n\nThe Skill walks through the 5-stage structure, runs a 10-item self-check,\nand warns you about IP names that may be blocked by Seedance 2.0.\n\n---\n\n## Compatible video models\n\nThe 5-stage structure is **model-agnostic**. Here's how the major 2026\nengines compare — single-shot ceiling, negative-prompt support, IP filter,\npreferred prompt language, and the one gotcha that trips people up:\n\n| Model | Max single shot | Negative prompt | IP filter | Lang | Notes \u002F gotcha |\n|---|---|---|---|---|---|\n| **Seedance 2.0** (Doubao \u002F Jimeng-Xiaoyunque, ByteDance) — *Mx-Shell's primary engine* | ~10–15s — but the **Doubao app is locked to 5s\u002F10s preset buttons**; the full 4–15s range only on Jimeng\u002FDreamina web + VolcEngine console | Partial — no reliable dedicated field in the consumer app; negate by describing what you *do* want | Strict — domestic platforms aggressively reject named celebrities + branded IP | Either (ZH native, EN works) | Duration depends entirely on the front-end. Don't promise 15s if the user is on the Doubao app. Native synced audio-video is its standout strength. |\n| **Veo 3 \u002F 3.1** (Google) | 8s per clip (4\u002F6\u002F8s); Extend adds 7s hops up to ~148s, but quality degrades after 4–5 extensions | Yes — dedicated field. List unwanted elements as plain nouns (`extra limbs, glitch morphs`); **no `no`\u002F`don't` command phrasing** | Strict — rejects public figures, brands, voice\u002Flikeness; scans prompt *and* frames | EN | The negative field wants descriptive phrases, not commands — `no rain` style instructions can backfire. Best-in-class native audio + realism. |\n| **Kling 2.x \u002F 3.0** (Kuaishou) | 2.5: 5–10s (Pro ~12s); **3.0: up to ~15s single-prompt** multi-shot | Yes — dedicated field. Use for stability artifacts (sliding feet, extra fingers, morphing), not generic \"quality\" words | Strict — a pre-gen banned-word filter rejects the **whole prompt** on one match; hypersensitive | Either (ZH native, EN strong; 3.0 multilingual dialogue) | The banned-word filter is notoriously over-sensitive — a benign body\u002Fcontact word can block a clean prompt. Sanitize wording first. Excellent action\u002Fmotion realism. |\n| **Hailuo \u002F MiniMax** (02 \u002F 2.3) | ~6–10s — 1080p caps ~6s, 768p extends ~10s | Yes — but vendor guidance says use sparingly, for specific artifacts not as a primary lever | Moderate — more permissive than Sora\u002FVeo, still blocks named celebs + overt IP | Either (ZH native, EN solid) | Resolution and duration trade off — you can't get max of both. Pick the axis that matters per shot. Strong motion at low cost. |\n| **Wan 2.x** (Alibaba, open-source) | 2.2: ~3–8s @ 24–30fps; 2.5\u002F2.6 extend ~10–15s by mode | Yes — robust field; defaults like `morphing, warping, face deformation, flickering` | Lenient — open-weights\u002Fself-hostable, so **no enforced filter when run locally** (hosted APIs may add their own) | Native ZH (both, but ZH-trained) | Leans Chinese — first\u002Flast-frame mode especially; EN-only prompts can underperform, add ZH for tricky shots. Self-hostable, full ComfyUI control, renders legible ZH\u002FEN on-screen text. |\n| **Runway Gen-4 \u002F 4.5** | 5s or 10s per generation | **No — does NOT support negatives.** `avoid X \u002F no X` can produce the opposite. Describe only what *should* appear | Strict — blocks celebs, real people, copyrighted characters\u002Fbrands | EN | Negative prompts are actively harmful here — `no distorted hands` can summon distorted hands. The single biggest mistake when porting prompts. Director-grade camera control + mature pro pipeline. |\n| **Pika** (2.2 \u002F 2.5) | Standard + Pikascenes: 5s or 10s; **Pikaframes (keyframe) up to ~25s** | Partial — 2.5 supports negatives (`no morphing, no extra limbs`); 2.2 unclear, verify in-app | Moderate — blocks overt celebs\u002FIP, generally more relaxed than Sora\u002FVeo | EN | Only the Pikaframes keyframe path reaches ~25s — ordinary text\u002Fimage-to-video is still 5s\u002F10s. Fast, effects\u002Ftransition-driven, great for stylized short-form. |\n| **Sora 2 \u002F 2 Pro** (OpenAI) | Up to ~25s continuous single-pass on Sora 2 Pro (standard tier shorter) | No dedicated field — negate in-prompt with guardrails like `original characters only, no logos or text` | Strict — triple-layer moderation; blocks named IP **and visual lookalikes** even without the name | EN | The filter catches *descriptions*, not just names — `a spiky-haired ninja in an orange jumpsuit` still trips it (matches Naruto). Avoid recognizable trait-bundles, not only proper nouns. Leading prompt comprehension + world coherence. |\n\n\u003Csub>Durations and negative-prompt mechanics for Veo 3.1, Runway Gen-4, Kling,\nWan, and Sora are consistent across multiple 2026 vendor\u002Fhelp-doc sources;\nSeedance 2.0 and Hailuo figures lean on third-party guides (treat `~` as\napproximate). \"Veo ~148s\", \"Sora\u002FPika ~25s\" come from extension\u002Fkeyframe\nfeatures, **not** plain single-shot generation. IP-filter \"strictness\" labels\nare qualitative.\u003C\u002Fsub>\n\n---\n\n## Sister projects (by the same author)\n\nThis is the video-prompt sibling of the AI-coding ecosystem maintained\nby [@jnMetaCode](https:\u002F\u002Fgithub.com\u002FjnMetaCode):\n\n- [superpowers-zh](https:\u002F\u002Fgithub.com\u002FjnMetaCode\u002Fsuperpowers-zh) — Chinese-enhanced edition of `obra\u002Fsuperpowers` (TDD \u002F debug \u002F git workflow skills)\n- [agency-agents-zh](https:\u002F\u002Fgithub.com\u002FjnMetaCode\u002Fagency-agents-zh) — 211 plug-and-play AI expert personas\n- [agency-orchestrator](https:\u002F\u002Fgithub.com\u002FjnMetaCode\u002Fagency-orchestrator) — Multi-agent collaboration orchestrator\n- [ai-coding-guide](https:\u002F\u002Fgithub.com\u002FjnMetaCode\u002Fai-coding-guide) — 66 Claude Code tips\n- [shellward](https:\u002F\u002Fgithub.com\u002FjnMetaCode\u002Fshellward) — AI Agent security middleware\n- [ai-coding-trilogy](https:\u002F\u002Fgithub.com\u002FjnMetaCode\u002Fai-coding-trilogy) — AI coding three-volume book\n\nAll projects share the same `SKILL.md` format. The video skill stacks\nfreely with any of them.\n\n---\n\n## Contributing\n\nThe roadmap is to collect more AI-shortfilm creators' methods, structured the\nsame way. New prompts, templates, fixes, and translations are welcome — see\n[CONTRIBUTING.md](.\u002FCONTRIBUTING.md) for the submission template and rules\n(public source required, credit the original creator).\n\nMade something with the method? It goes on the [showcase](.\u002Fshowcase.md).\n\n---\n\n## License\n\n**[MIT License](.\u002FLICENSE)** for everything authored by jnMetaCode\n(methodology, FAQ, templates, Skill, metadata). Free for any use\nincluding commercial — just keep the copyright notice.\n\nMx-Shell's original prompts and document excerpts — sourced from his\nfan-group documents and public Douyin livestream that he himself\ndistributed — remain **© Mx-Shell, all rights reserved**. Archived here\nfor educational reference; commercial use requires contacting Mx-Shell\ndirectly. Full dual-licensing details: [NOTICE](.\u002FNOTICE) ·\nattribution: [credits.md](.\u002Fcredits.md).\n\n---\n\n## A line worth remembering\n\n> *\"For creation, the equipment is not what matters.\n> The idea is what matters.\"*\n> — Mx-Shell, May 12, 2026 livestream\n","该项目提供了制作高质量AI短片的方法论、提示库及Claude Code技能，背后支撑的是被好莱坞导演PJ Ace誉为“近年来最佳短片之一”的《僵尸拾荒者》。核心功能包括详细的制作流程、丰富的文本到视频提示以及与多个AI视频生成平台（如Sora、Seedance等）的兼容性，利用Python语言实现。特别适合于希望使用AI技术创作短视频内容的创作者，无论是专业电影人还是业余爱好者都能从中受益，快速上手并产出令人印象深刻的视觉作品。",2,"2026-06-11 04:10:18","CREATED_QUERY"]