Skip to content

norahe0304-art/30x-video / cli/skill

30x-web-to-video

Flagged

Use when a user gives a brand URL and wants a premium launch, demo, or marketing video. Zero-manual workflow — Claude runs the V2 orchestrator itself (scrapes brand evidence, auto-installs missing tools, generates the first-cut Remotion project). User only sees the URL and the Studio preview.

FileKindSize
.gitignoreother103 B
AGENTS.mdmarkdown2.5 KB
FUSION-REPORT.mdmarkdown12.3 KB
README.mdmarkdown3.6 KB
SKILL.mdskill md30.7 KB
benchmarks/AGENTS.mdmarkdown773 B
benchmarks/README.mdmarkdown2.8 KB
benchmarks/manifest.jsonconfig5.9 KB
examples/INDEX.mdmarkdown1.8 KB
examples/caylent/MainVideo.tsxother6.0 KB
examples/caylent/scenes-v2/Act1V2.tsxother1.9 KB
examples/caylent/scenes-v2/Act2V2.tsxother2.5 KB
examples/caylent/scenes-v2/Act3V2.tsxother11.3 KB
examples/caylent/scenes-v2/Act4V2.tsxother8.6 KB
examples/caylent/scenes-v2/Act5V2.tsxother1.0 KB
examples/caylent/scenes/Act1.tsxother6.0 KB
examples/caylent/scenes/Act2.tsxother2.5 KB
examples/caylent/scenes/Act3.tsxother7.4 KB
examples/caylent/scenes/Act4.tsxother7.1 KB
examples/caylent/scenes/Act5.tsxother2.2 KB
examples/caylent/theme.tsscript1.2 KB
examples/cofounder/MainVideo.tsxother5.5 KB
examples/cofounder/scenes/AGENTS.mdmarkdown1.4 KB
examples/cofounder/scenes/Act1.tsxother2.3 KB
examples/cofounder/scenes/Act2.tsxother2.4 KB
examples/cofounder/scenes/Act3a.tsxother6.7 KB
examples/cofounder/scenes/Act3b.tsxother8.5 KB
examples/cofounder/scenes/Act3c.tsxother6.8 KB
examples/cofounder/scenes/Act4.tsxother4.4 KB
examples/cofounder/scenes/Act5.tsxother3.3 KB
examples/cofounder/scenes/World.tsxother4.0 KB
examples/cofounder/theme.tsscript1.5 KB
examples/happy-model/MainVideo.tsxother6.8 KB
examples/happy-model/scenes/AGENTS.mdmarkdown1.7 KB
examples/happy-model/scenes/Act1.tsxother7.9 KB
examples/happy-model/scenes/Act2.tsxother10.2 KB
examples/happy-model/scenes/Act3.tsxother15.6 KB
examples/happy-model/scenes/Act4.tsxother16.9 KB
examples/happy-model/scenes/Act5.tsxother9.0 KB
examples/happy-model/scenes/copy.tsscript588 B
examples/happy-model/theme.tsscript1.2 KB
examples/mana/MainVideo.tsxother4.9 KB
examples/mana/scenes/AGENTS.mdmarkdown926 B
examples/mana/scenes/Act1Open.tsxother2.8 KB
examples/mana/scenes/Act2Hero.tsxother3.3 KB
examples/mana/scenes/Act3Flavors.tsxother9.0 KB
examples/mana/scenes/Act4Label.tsxother7.3 KB
examples/mana/scenes/Act5Close.tsxother4.7 KB
examples/mana/scenes/World.tsxother7.2 KB
examples/mana/theme.tsscript2.6 KB
examples/parker/MainVideo.tsxother6.2 KB
examples/parker/scenes/AGENTS.mdmarkdown1.1 KB
examples/parker/scenes/Act1Hook.tsxother3.2 KB
examples/parker/scenes/Act2Meet.tsxother2.3 KB
examples/parker/scenes/Act2Problem.tsxother3.6 KB
examples/parker/scenes/Act3Steps.tsxother4.7 KB
examples/parker/scenes/Act4Proof.tsxother3.2 KB
examples/parker/scenes/Act5Close.tsxother2.6 KB
examples/parker/scenes/PaperWorld.tsxother6.8 KB
examples/parker/theme.tsscript1.6 KB
examples/perplexity-pc/MainVideo.tsxother4.5 KB
examples/perplexity-pc/scenes/Act1.tsxother2.1 KB
examples/perplexity-pc/scenes/Act2.tsxother2.7 KB
examples/perplexity-pc/scenes/Act3.tsxother3.1 KB
examples/perplexity-pc/scenes/Act4.tsxother1.6 KB
examples/perplexity-pc/scenes/Act5.tsxother2.0 KB
examples/perplexity-pc/theme.tsscript2.5 KB
references/AGENTS.mdmarkdown767 B
references/animations.mdmarkdown15.1 KB
references/audio.mdmarkdown8.6 KB
references/components.mdmarkdown9.5 KB
references/lottie.mdmarkdown5.1 KB
references/visual-effects.mdmarkdown5.0 KB
rules/AGENTS.mdmarkdown3.0 KB
rules/archetypes.mdmarkdown6.0 KB
rules/artifact-catalog.mdmarkdown10.8 KB
rules/beat-sync.mdmarkdown5.7 KB
rules/cards.mdmarkdown2.5 KB
rules/cinematic.mdmarkdown6.7 KB
rules/color.mdmarkdown2.2 KB
rules/composition.mdmarkdown13.2 KB
rules/creative-moves.mdmarkdown3.8 KB
rules/data-viz.mdmarkdown1.9 KB
rules/finish-gate.mdmarkdown5.2 KB
rules/layout.mdmarkdown6.6 KB
rules/media-resolve.mdmarkdown7.1 KB
rules/motion.mdmarkdown9.5 KB
rules/narration-sync.mdmarkdown5.5 KB
rules/narrative-templates.mdmarkdown10.4 KB
rules/narrative.mdmarkdown7.3 KB
rules/qc-gates.mdmarkdown11.3 KB
rules/taste.mdmarkdown12.6 KB
rules/transitions.mdmarkdown2.7 KB
rules/typography.mdmarkdown4.1 KB
rules/ui-mockups.mdmarkdown5.6 KB
rules/workflow.mdmarkdown53.2 KB
scaffold/.prettierrcother66 B
scaffold/AGENTS.mdmarkdown1.1 KB
scaffold/eslint.config.mjsscript79 B
scaffold/package-lock.jsonmanifest186.3 KB
scaffold/package.jsonmanifest1.5 KB
scaffold/remotion.config.tsscript219 B
scaffold/src/AGENTS.mdmarkdown831 B
scaffold/src/MainVideo.tsxother41.7 KB
scaffold/src/Root.tsxother2.3 KB
scaffold/src/components/AGENTS.mdmarkdown1.5 KB
scaffold/src/components/Animations.tsxother6.8 KB
scaffold/src/components/Atmosphere.tsxother8.1 KB
scaffold/src/components/Background.tsxother4.5 KB
scaffold/src/components/Captions.tsxother6.4 KB
scaffold/src/components/LogoEntrance.tsxother15.2 KB
scaffold/src/components/MockupTemplates.tsxother19.0 KB
scaffold/src/components/PhoneFrame.tsxother3.6 KB
scaffold/src/components/ShaderFX.tsxother5.3 KB
scaffold/src/components/Transitions.tsxother8.4 KB
scaffold/src/components/UI.tsxother3.0 KB
scaffold/src/components/VisualEffects.tsxother10.7 KB
scaffold/src/generated/AGENTS.mdmarkdown508 B
scaffold/src/generated/beat-map.tsscript1.1 KB
scaffold/src/generated/project-data.tsscript5.2 KB
scaffold/src/index.cssother576 B
scaffold/src/index.tsscript400 B
scaffold/src/lib/AGENTS.mdmarkdown411 B
scaffold/src/lib/fonts.tsscript867 B
scaffold/src/lib/modes.tsscript2.2 KB
scaffold/src/lib/utils.tsscript352 B
scaffold/src/theme.tsscript2.5 KB
scaffold/tsconfig.jsonconfig333 B
scripts/AGENTS.mdmarkdown2.4 KB
scripts/analyze-audiomap.pyscript13.3 KB
scripts/beat-sync.tsscript12.1 KB
scripts/benchmark-suite.tsscript17.8 KB
scripts/critique-scenes.tsscript10.1 KB
scripts/evidence-model.tsscript9.4 KB
scripts/generate-bgm.pyscript10.5 KB
scripts/iterate.tsscript4.0 KB
scripts/project-blueprint.tsscript12.1 KB
scripts/regenerate-scene.tsscript7.4 KB
scripts/render-qa.tsscript9.0 KB
scripts/scene-constitution.tsscript13.4 KB
scripts/timing-audit.tsscript9.5 KB
scripts/url-to-video.tsscript68.6 KB
scripts/visual-audit.tsscript26.5 KB

rules/narration-sync.md

View on GitHub
<!--
[INPUT]: TTS 生成的 narration 音频 (可选路径), whisper 词级时间戳 JSON, STORYBOARD/story.md 的 beat 计划, scene 时长表
[OUTPUT]: 三条叙述同步法则 — Beat-Lock (视觉变化锁词级时间戳)、Caption/Brand Cleanup (字幕清理不动时间戳)、Sentence-Boundary Snapping (切点吸附句界)
[POS]: rules/ 的配音同步层; 仅当视频含 voiceover/captions 时生效 (纯 BGM 视频跳过); 被 SKILL.md Step 4 引用, 披露项汇入 qc-gates.md Gate 4
[PROTOCOL]: 变更时更新此头部,然后检查 AGENTS.md
-->

# Narration Sync — Beat-Lock, Caption Cleanup, Boundary Snapping

Applies ONLY when the video has a voiceover (TTS narration) or burned captions. BGM-only videos skip this file entirely. When narration exists, these three rules are mandatory — timing "estimated from the script" is the #1 way narrated videos ship broken.

## Rule 1: Beat-Lock — pin every visual change to real word timestamps

Never assume timing. A 78-word script is not "about 30 seconds" — it is exactly as long as the rendered audio says it is, and each sentence lands where whisper says it lands.

After generating the TTS audio:

```bash
whisper public/brand/narration.mp3 --model base --word_timestamps True --output_format json --output_dir evidence/
```

(Any whisper flavor works — what matters is a JSON of `{word, start, end}` per word. faster-whisper and whisper.cpp emit equivalents.)

Then:

1. Convert word times to frames (`t × fps`, round) and derive each act's real start/end from where its narration sentences actually sit.
2. Every scene cut, headline reveal, count-up start, and caption cue in `MainVideo.tsx` derives from these numbers — not from an even split of `durationInFrames`, not from reading the script and eyeballing.
3. If the audio runs long/short vs the plan, update the scene constitution's frame ranges — the audio is the source of truth, the plan is the draft.

**Forbidden:** writing `<Sequence from={900}>` because "act 3 should start around 30s". If a frame number can't be traced to a word timestamp or a beat-map entry, it's a guess.

**Gate:** for a narrated video, every act boundary in MainVideo.tsx maps to a word/sentence timestamp from the whisper JSON (state the mapping in the act's Gate-2 evidence block).

## Rule 2: Caption & brand-word cleanup — fix text, never touch timestamps

Raw whisper output is not caption-ready. Before rendering captions, Claude passes over the transcript:

1. **Fix brand-word mis-transcriptions** from surrounding context — whisper writes "Kalent" for "Caylent", "strip" for "Stripe". Cross-check every product/brand token against `brand-report.json`'s brandName.
2. Remove filler words (um, uh, you know) from caption *text* only.
3. **Keep all timestamps unchanged.** Edit only the text field; the word timing grid is what Rule 1 and Rule 3 depend on.
4. Write the cleaned copy to a separate file (`evidence/transcript-clean.json`), never over the raw transcript.

**Disclosure hook (→ Gate 4):** a mis-transcribed brand name is also a *pronunciation warning* — whisper may be spelling what the TTS voice actually said. You cannot verify pronunciation without ears. Add to "What I did NOT verify": "whisper heard '<X>' for '<Brand>' — listen to the first 5s; if mispronounced, regenerate TTS with a phonetic spelling."

## Rule 3: Sentence-boundary snapping — cuts land on sentence ends, never mid-word

Scene transitions in a narrated video must land in the gaps between sentences:

1. From the whisper JSON, a sentence boundary = a word whose text ends in `.?!`; the cut point is that word's `end` + ~300ms pad, converted to frames.
2. If a planned cut falls mid-sentence, move it to the nearest sentence end (prefer extending; a beat that breathes beats a beat that clips a clause).
3. When BGM beat-sync (`beat-map.json`) is also active, narration wins: pick the BGM beat *nearest to* the sentence gap, not the reverse — a cut on a musical beat but mid-word is still a broken cut.
4. Cross-check with silence: `ffmpeg -i narration.mp3 -af silencedetect=noise=-35dB:d=0.3 -f null -` — a real pause near your cut point confirms it; no pause anywhere near it means the sentence isn't over.

**Gate:** no act transition occurs while a narration word is mid-utterance (word `start < cut < end` for any word = fail).

## Rule 4: Continuous speech — 一段论证一口气说完, 画面吸附短语

用户判决(2026-07-04):为了给每个卖点配画面而把一段旁白拆成几个短句文件、句间留 1-2s 死气口 = "说话断掉了,很不舒服"。同一段论证(如证明段的三个卖点)必须是**一段连续生成的音频**,中途不许有人为空档。

对齐方向是单向的:**画面跟话走,话不迁就画面。**

1. 把整段写成一份 script 连续生成 TTS;用词级对齐(ElevenLabs `with-timestamps` 直接返回,或 whisper)量出每个卖点短语的 onset。
2. 段内视觉切点(beat 边界常量)设为短语 onset 对应的帧 — 不是先定画面时长再找话塞。
3. 卖点节奏太快的正解是**加长文案给每个点更多词**,再按新 onset 重吸附 — 不是拆音频。
4. 有旁白时,headline 下的解释性副句小字一律删("干净一点")— 信息由旁白承载,画面只留巨字排 + 世界本身。产品世界原生的机器文本(终端行、状态页区域标签、`ms` 延迟戳)不算副句,保留。

**Gate:** 同一论证段的旁白音频文件数 = 1;每个段内 beat 边界能指认到对齐 JSON 里的短语 onset。
planning-and-task-breakdownaddyosmani/agent-skillsBreaks work into ordered tasks. Use when you have a spec or clear requirements and need to break work into implementable tasks. Use when a task feels too large to start, when you need to estimate scope, or when parallel work is possible.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
screenpipethedotmack/claude-memSearch screen and audio history recorded by Screenpipe and optionally save selected context into claude-mem. Use when the user asks to recall desktop activity, a meeting, or something seen in another app through Screenpipe.Stars: 118kUpdated Claude CodeCodexVerifiedInstall
contact-widgetnexu-io/open-designSelf-contained floating chat widget with welcome screen, social links, meeting button, and message input. Single HTML file, zero dependencies.Stars: 108kUpdated Claude CodeCodexVerifiedInstall
meeting-notesnexu-io/open-designMeeting notes page — title bar with attendees, agenda checklist, decisions block, action items table with owners + dates, and a "next meeting" footer. Use when the brief mentions "meeting notes", "minutes", "1:1 notes", "all-hands recap", or "会议纪要".Stars: 108kUpdated Claude CodeCodexVerifiedInstall