LibTV Video Production (Storyboard to Short Film)

VerifiedCaution

Create short videos from storyboards using LibTV CLI. Covers full pipeline: 9-grid storyboard, image2video generation, ffmpeg editing, prompt templates, and common pitfall solutions.

Sby Skills Guide Bot
ContentAdvanced
7/24/2026
Claude CodeCodexCursor
#video-generation#storyboard#seedance#ffmpeg#prompt-templates

Recommended for

Automated summary

End-to-end methodology for turning a storyboard sheet into a short film with the LibTV CLI: Seedance 2.0 image-to-video generation, cleanup of annotations on the panels, then local editing with ffmpeg (transitions, score, narration).

Strengths

  • Detailed pipeline from start to finish: nine-panel storyboard, per-shot image2video at 720p then 1080p, local ffmpeg assembly.
  • Reusable prompt recipe library (style lock, character lock, body-proportion lock, zero text, locked scene, positive enumeration of on-screen elements).
  • Pitfall table with root cause and fix: character drift, panel numbers leaking into frames, extra entities, safety-review rejections.
  • Hands-on environment notes: inline PowerShell for Chinese prompts, non-ASCII Windows paths, manual URL polling when a 1080p job hangs at 99%.

Limitations

  • Heavily tied to the proprietary LibTV/Seedance tooling and CLI, including a mandatory web login.
  • The SKILL.md constantly defers to AGENTS.md and references/*.md, which are not included here, so the file alone is partial.
  • Instructions are written for Windows/PowerShell; macOS and Linux equivalents are only sketched for skill installation.
When to use it

Use it when you want to build a short video from storyboard frames or a script with liblib.tv / LibTV / Seedance, or maintain such a production project.

When not to use it

Avoid it for regular video editing without AI generation, or when you cannot use the LibTV CLI or generate images and videos with AI.

Security analysis

Caution
Quality score84/100

The file is a methodology/documentation skill for a licensed CLI video-production tool. It gives concrete PowerShell/Bash commands that call a local binary and cloud APIs and run ffmpeg, which is powerful-tool usage but clearly aligned with the stated purpose, and it even requires user sign-off before paid generation runs. Nothing destructive, obfuscated, or exfiltrating; the only sensitive artifact is the tool's own credential file, which is referenced but never transmitted elsewhere.

Findings
  • •Instructs invoking a locally installed third-party binary (~/.libtv/libtv.exe) and authenticating via `libtv login web`, which stores credentials in ~/.libtv/credentials.json — credentials are not exfiltrated, but the agent will handle an auth token file.
  • •Uses remote AI generation APIs (Seedance / nebula / seedream) and ffmpeg locally; both are network/execution-capable tools, used for the stated legitimate purpose.
  • •Explicitly asks the agent to confirm the plan with the user before running video generation (`--run`), which is a good guardrail rather than a risk.
  • •No destructive commands, no curl|sh, no base64 payloads, no instructions to disable safety or send data to third-party endpoints.
How this analysis is made

Examples

Storyboard to short film
帮我把这 9 张分镜图用 libtv 做成一支 60 秒短片,2D 绘本风格,先 720p 试错
Single shot image2video
用 Seedance 把 sb_07_clean 生成 8 秒视频,角色定版图放第一张参考,16:9、720p,prompt 带上角色铁律和零文字铁律
Assembly with ffmpeg
所有镜头都齐了,用 ffmpeg 把 ~/proj/shots 里的片段剪成一版:加转场、配乐和旁白

name: libtv-video-production description: 用 LibTV CLI(Seedance 2.0 视频 + nebula/seedream 图)把分镜做成短片(2D 绘本或写实皆可)——涵盖九宫格故事板→clean 去标注→image2video 单镜生成→ffmpeg 本地剪辑(转场+配乐+旁白)全流程,附 prompt 铁律模板库(角色/比例/零文字/场景锁定/负面词反召唤)与高频踩坑速查(多狗/角色漂移/格号穿帮/审核三连拒/长镜头漂移)。当用户要用 liblib.tv / libtv / Seedance 把参考图或脚本做成视频短片、或维护此类制作工程时触发。

LibTV 短片制作(分镜 → 成片)

把"一组分镜"用 LibTV CLI 做成一支短片的通用方法论。美术基调可换(2D 绘本 / 写实 / 任意风格),方法论不变。本 SKILL.md 只给精简操作指引 + 速查;完整细节看同目录 AGENTS.md 与 references/*.md(见末尾索引)。

⚠️ 旧 sk-libtv-* OpenAPI 已废弃,只用 CLI(libtv login web 登录)。旧 API 建的项目不进 web dashboard。


心智模型(一句话)

定调(质感词) + 建定版资产(角色/场景/道具) + 画故事板九宫格
  → 每镜 image2video(挂 故事板+定版图) 720p 试错
  → 全部齐 → ffmpeg 本地剪辑(转场+配乐+旁白) → 1080p 定稿
  • CLI 三层结构:项目 → [分组] → 节点。节点参数 -s 写 data.params(走 schema),-u 写节点顶层属性。
  • 故事板 = 唯一拍摄脚本:九宫格只决定"怎么拍/表达什么"(构图/景别/运镜/表演),风格色彩由设定图 + 质感词承担。
  • 剪辑在本地做,不靠云端 video-clip 节点——ffmpeg 对转场/配乐/字幕可控得多。

SOP(每镜标准流程)

  1. 看故事板九宫格实际画面,决定本镜挂哪些定版参考图(不按旧剧本文字加戏)。
  2. 挑参考图:角色定版图放第一张(锚脸/发型)+ 场景图 + 道具图(用 in-scene 合成图更稳)。
  3. 写 prompt:质感词 + 按需套铁律模板(角色/比例/零文字/场景锁定/装备)。
  4. 生成视频前先把方案报给用户,得到"OK"才 --run(图片生成可直接做)。
  5. 720p 试错(便宜),过关后再 1080p 定稿。
  6. 段尾留转场钩子(光/门/手等出口动作),方便后期融合。
  7. 全部镜头齐 → ffmpeg 统一剪辑成片。

单镜命令骨架:

$libtv = "$env:USERPROFILE\.libtv\libtv.exe"
& $libtv node create "shot07_video" -t video `
  --prompt "{{Image 1}} 锚定脸型发型;<本镜画面描述> + <质感词> + <按需铁律>" `
  --left "wife_fixed" --left "sb_07_clean" --left "scene_classroom" `
  -s model=star-video2 -s modeType=image2video `
  -s ratio=16:9 -s resolution=720p -s duration=8 -s enableSound=off --run

关键铁律(5–8 条,违反多半返工)

| # | 铁律 | 要点 | |---|---|---| | 1 | 质感词锁死、每镜必带 | 开工先定一句固定质感词(如 2D 绘本:2D柔和动画绘本插画风格…gentle 2D storybook;绝不要3D/写实),所有图/视频原样复用,保证全片同风格。 | | 2 | 定版图作锚、角色图放第一张 | 凡涉及已定版的角色/场景/道具,必须挂定版图,绝不只靠文字还原。九宫格重绘用 4K(2K 小格脸糊)。 | | 3 | 【角色铁律】+【人物比例铁律】 | 严格按第一张图,9 格同脸同发型同服装;写明年龄+成年约 6.5 头身,禁幼态/大头娃娃/裁头/变形。 | | 4 | 【零文字零数字铁律】 | 格内不写任何字(单据画横线、屏幕画亮屏);prompt 注明"格号/箭头是标注不是画面内容",否则视频会照抄穿帮。 | | 5 | 负面词不点名(反召唤) | Seedance 负面指令弱:要排除某元素就通篇不提它的名字,改用【画面元素铁律】正面穷举"画面里有且只有:…"("别想大象"效应)。 | | 6 | 【场景锁定铁律】+ 控时长 | 长镜头后半会漂移:写"全程同一房间、家具按图不变";时长压到 ≤12s 降漂移。 | | 7 | 【镜头铁律】以人为主体 | 挂场景图易变成环境展示片:写"九宫格是唯一脚本、镜头紧跟人物、禁长段空镜/环境运镜,场景只是背景"。 | | 8 | 真实记忆镜头先问用户 | 含真实仪式/道具/构图的镜头别凭脚本猜,先拿照片或口述参照再设计。 |

铁律完整模板(可直接复制)见 references/prompt-recipes.md。


高频坑速查(最常踩的 3–4 个)

| 坑(案例) | 根因 | 解法 | |---|---|---| | 角色漂移/换脸(案例 06 长发漂移、13–16 幼态裁头) | 编辑九宫格没挂锚点 / 分辨率低 | 重绘而非编辑:定版图放第一参考 + 4K + 角色/比例铁律 | | 格号/文字被照抄进画面(案例 19/22/27 气球上"②"、20 账单文字) | 九宫格里有格号、箭头、文字 | 零文字铁律;仍穿帮就 nebula image2image 做 sb_XX_clean(去格号/箭头)喂片版,clean 版也要逐格 QC(重绘可能加戏) | | 多出实体(案例:第 3 只狗、水下凭空冒出狗) | 参考图本身含该元素 / 负面词点名召唤 | 先查所有参考图有没有那元素,再改 prompt;用干净参考 + 正面穷举画面元素,排除项不点名 | | 审核"未通过安全审核"三连拒(案例 30) | 某参考图或某格触发审核(如人脸贴宠物特写) | 5s 试探片 + 参考图二分排查(约 1 分钟出结果)定位真凶后替换;积分会自动退 |

全量 33+ 坑(CLI 操作 / 一致性 / 长镜头 / 道具 / 剪辑 / 审核 / Windows 中文路径)见 references/pitfalls.md。


Windows / 环境要点(一次性)

$libtv = "$env:USERPROFILE\.libtv\libtv.exe"   # CLI 路径
& $libtv login web                              # 浏览器登录,凭据落 ~/.libtv/credentials.json
# cd 到项目目录后:libtv project use <uuid> 绑定 cwd
  • 中文 prompt 用内联 PowerShell,不写 .ps1(GBK 读乱码)。
  • cv2 等读中文路径用 imdecode(np.fromfile(...)) / imencode+tofile。
  • 1080p/长视频 --run 可能卡 99% 报 failed 但任务在跑:提交后单独轮询节点 data.url 兜底下载。
  • 音乐节点 create 报"场景不一致"→ 加 -s scene=Music。

详见 references/content-safety-and-env.md。


安装方式

把本仓库目录(含 SKILL.md + AGENTS.md + references/)复制或软链到 Claude Code 技能目录即可被自动发现。<REPO_DIR> 替换为你 clone 本仓库后的本地绝对路径:

# Windows(软链推荐,改一处两处同步)
New-Item -ItemType SymbolicLink `
  -Path "$env:USERPROFILE\.claude\skills\libtv-video-production" `
  -Target "<REPO_DIR>"
# macOS/Linux 等价
ln -s "<REPO_DIR>" ~/.claude/skills/libtv-video-production

目录结构须保持 SKILL.md 在技能根、references/ 为同级子目录,否则下方索引的相对路径会失效。其他工具(Codex / CodeBuddy 等)的接入见仓库根 README.md。


完整细节索引(读这些拿全文)

| 想做什么 | 看这个文件 | |---|---| | 完整方法论(前期定调→建资产→生成→剪辑全流程深讲) | AGENTS.md | | CLI 全命令矩阵 / 节点类型 / -s vs -u / NDJSON 管道 / 能力边界 | references/cli-reference.md | | 端到端制作工作流(三阶段每步怎么落地) | references/production-workflow.md | | Prompt 铁律模板全文(角色/比例/零文字/场景锁定/装备/in-scene 锚定…可复制) | references/prompt-recipes.md | | ffmpeg 本地剪辑流水线(转场 xfade / 音轨 acrossfade / 字幕 / 拼装核对) | references/editing-pipeline.md | | 内容审核排查 + 真人合规 + Windows/PowerShell 怪癖 | references/content-safety-and-env.md | | 全量踩坑速查表(按主题分类,跨项目通用) | references/pitfalls.md |

From the file to a first result

The download contains exactly the content displayed on this page. This copy may differ from the latest version in the source repository.

View the source and supporting files

This download contains SKILL.md only. Any scripts, references, templates and dependencies are not included. Get the required files from the source, preserving their directory structure, before continuing.

1. Download the file

Download SKILL.md

From the root of a test project, save the file as .skill-downloads/libtv-video-production-storyboard-to-short-film-mryiqiuy/SKILL.md, or use the command below. This folder is for inspecting the skill before installation.

Command for macOS/Linux or WSL: requires a POSIX shell and curl. On Windows without WSL, use the download button and create the folder manually.

mkdir -p '.skill-downloads/libtv-video-production-storyboard-to-short-film-mryiqiuy' && curl --fail --show-error --location 'https://skills-guides.com/api/skills/libtv-video-production-storyboard-to-short-film-mryiqiuy/download' --output '.skill-downloads/libtv-video-production-storyboard-to-short-film-mryiqiuy/SKILL.md.part' && mv '.skill-downloads/libtv-video-production-storyboard-to-short-film-mryiqiuy/SKILL.md.part' '.skill-downloads/libtv-video-production-storyboard-to-short-film-mryiqiuy/SKILL.md'

2. Check prerequisites

Read SKILL.md before using it. If it references scripts or external services, inspect them and prepare the required access in your test environment.

No recognized prerequisites section in this file. This does not mean it is self-contained: libraries, tools and access requirements may be mentioned elsewhere in the content or source repository.

The platforms on this page are catalog indications. The paths below follow tool documentation; they do not prove this skill has been run in those tools. No minimum tool version is confirmed for this listing.

3. Load the skill

After inspection, copy the complete folder into your project using one of the commands below, then open a new tool session. The command stops if a folder with the same name already exists: compare the files before replacing it.

Claude Code
test -f '.skill-downloads/libtv-video-production-storyboard-to-short-film-mryiqiuy/SKILL.md' && test -s '.skill-downloads/libtv-video-production-storyboard-to-short-film-mryiqiuy/SKILL.md' && mkdir -p '.claude/skills' && test ! -e '.claude/skills/libtv-video-production' && cp -R '.skill-downloads/libtv-video-production-storyboard-to-short-film-mryiqiuy' '.claude/skills/libtv-video-production'

Cursor
test -f '.skill-downloads/libtv-video-production-storyboard-to-short-film-mryiqiuy/SKILL.md' && test -s '.skill-downloads/libtv-video-production-storyboard-to-short-film-mryiqiuy/SKILL.md' && mkdir -p '.cursor/skills' && test ! -e '.cursor/skills/libtv-video-production' && cp -R '.skill-downloads/libtv-video-production-storyboard-to-short-film-mryiqiuy' '.cursor/skills/libtv-video-production'

Codex
test -f '.skill-downloads/libtv-video-production-storyboard-to-short-film-mryiqiuy/SKILL.md' && test -s '.skill-downloads/libtv-video-production-storyboard-to-short-film-mryiqiuy/SKILL.md' && mkdir -p '.agents/skills' && test ! -e '.agents/skills/libtv-video-production' && cp -R '.skill-downloads/libtv-video-production-storyboard-to-short-film-mryiqiuy' '.agents/skills/libtv-video-production'

OpenSkills

OpenSkills must already be installed. The local install/read procedure is checked with OpenSkills 1.5.0, which requires Node.js ≥ 20.6.0. Catalog slugs are not resolved automatically.

Official OpenSkills installation
openskills install './.skill-downloads/libtv-video-production-storyboard-to-short-film-mryiqiuy'

openskills read 'libtv-video-production-storyboard-to-short-film-mryiqiuy'

Reading should display the installed file content and its base directory. It does not run the skill or install its dependencies.

For an agent that reads AGENTS.md, then sync the index. This command changes AGENTS.md: review the diff before using it.

openskills sync

Manual method or another tool: attach SKILL.md to a conversation, or open it in an agent that can read files. Include any required supporting resources. No universal folder is recognized by every tool.

4. Try a first task

Use a short, synthetic input: an API endpoint, a small document or an interface brief. In a new conversation, attach the file or give the agent its path, then adapt this request:

Read .skill-downloads/libtv-video-production-storyboard-to-short-film-mryiqiuy/SKILL.md and apply its instructions to this example: [describe a test input]. Identify any missing files, dependencies or access. Produce the result requested by the skill without inventing data or running unreviewed scripts.

Expected result: output matching the skill instructions, or a precise list of missing prerequisites. A read confirmation alone does not validate a first use. If an external resource is unavailable, that path remains untested.

Related skills