Video & Animation

Motion design foundation - timing, easing, enter/exit patterns, accessibility, performance.

Switch this session to Dudu (דודו), the AI promo video director, until the owner leaves Dudu mode.

Analyze a short-form video (TikTok, Instagram Reel, YouTube Short, or any MP4) to extract its full transcript, understand why it performed, and generate concrete repurposing ideas. Use this skill whenever the user asks to: analyze a video, understand why something went viral, reverse-engineer a creator's content, extract a transcript from a video, figure out what makes a video perform, or get repurposing ideas from a video. Also trigger when the user says things like "why did this perform", "what's the hook here", "can I repurpose this", "analyze this TikTok/Reel", or drops a video file path or URL.

Used by Dudu (AI promo video director) after the picture cut. Designs the sound (cue sheet synced to cuts), generates music, effects and voice (Gemini, TTS, recordings, free libraries), and mixes it onto the cut with tools/mix_audio.py into a finished film.

Used by Dudu (AI promo video director) for stages 3-5. Builds the style and character bible, the 12-shot list, reference sheets, location frames and the 12 keyframe prompts for Gemini / Nano Banana or ChatGPT / GPT Image, and reviews every generated image.

Used by Dudu (AI promo video director) to learn from a reference video link (YouTube and most video sites). Downloads it for private analysis, detects every cut, reviews one frame per shot, writes a structured analysis into library/study/<id>/analysis.md and updates library/INDEX.md, then deletes the video. Use when the owner shares a link and says "learn from this", "study this", "do it like this", or asks for a style taken from an ad.

Generate context-corrected English–Simplified Chinese subtitles for local English videos and optionally create a new JianYing draft. Use when a user wants transcription plus semantic correction, natural Chinese localization, bilingual SRT delivery, or a JianYing project with bilingual captions. Do not use for simple subtitle-format conversion when no transcription, correction, or translation is needed.

Make or continue a short animated promo video — runs every stage in order and resumes where the video left off.

Render a promo-clip video to MP4 in every chosen format and size, and QA the results.

Guided interview for a promo-clip video — writes brief.md, the theme and the output settings.

Write the animated scenes of a promo-clip video for its primary format — in parallel, one agent per scene, or one by one — then review the whole film and proof it with stills.

Add the other aspect-ratio layouts (16:9, 9:16, 1:1) to every scene of a promo-clip video — in parallel, one agent per scene, or one by one — and proof them.

>- 制作「数字人全程口播解说 + 真实案例展示」风格的技能/产品推广短视频（16:9，60-120s）： 所有场景人物开口说话（音频驱动口型同步，双引擎自动决策——短段 H3 音频驱动 / 长段 Wan2.2 S2V 720p）、人物服饰与案例提示词全参数化（换一份 config 即换人设/换案例）、 Blender 数据长廊确定性运镜（开场推进/收尾拉远，图表面板贴真实图表）、构图卡文字 直出（逐行核对清单+逐字 OCR 复验）、真实数据图表页、TTS 旁白、品牌 signature 水印、 三步式 ffmpeg 合成、五道 QC 闸门（文字稳定性/口型运动/手部逐指/音画对齐/FunASR 听写）。管线模板化，换一份 config 即可为任何 skill 出同款风格视频。当用户说 "给 XX skill 做推广视频/宣传片/爆款视频/数字人视频/口播讲解视频"，或要求复用 数字人主讲风格的 showcase 视频时使用。

Used by Dudu (AI promo video director) at every gate. Checks script claims against project/product.md, structure and pacing, and reviews images and clips with keep/fix/regenerate decisions into project/qa-notes.md.

Used by Dudu (AI promo video director) for stages 1-2. Interviews the owner for the brief and writes a 12-shot promo script (problem-agitate-solve, about 70/30 problem/solution, minimal dialogue, one CTA) into project/brief.md and project/script.md.

Used by Dudu (AI promo video director) for stage 6. Picks the animation model with a one-shot A/B test, writes image-to-video motion prompts for Kling (and Veo / Seedance), and reviews every clip with contact sheets and a motion report.

Obtain SRT from YouTube or Bilibili links using platform captions or automatic local Whisper transcription, then summarize, analyze, or answer questions with timestamps. Use for video subtitle and content-analysis requests; exclude general API questions and project development tasks.

Update a promo-clip video's engine copy to the plugin's current engine and compare the scenes before and after.

Used by Dudu (AI promo video director) to drive Gemini (Nano Banana images, Veo video), ChatGPT (GPT Image) and Kling (image-to-video) in a Playwright browser the owner logged into. Contains the working click paths, upload and download recipes, and the fixes for every failure we hit.

Used by Dudu (AI promo video director) for stage 7. Measures where each clip actually moves, cuts every shot to its action, color grades problem vs solution, and assembles the rough cut with ffmpeg via tools/build_cut.py.

Write the voice-over script for a promo-clip video, split into scenes with a picture idea each.

Make the characters, backdrops, props and screens for a promo-clip video — AI-generated in the house style or processed from your own visuals.

Craft reference for promo-clip videos — the scene API, per-format layout, word-timed events, AI provider quirks, QA thresholds and theme fields. Use when writing or reviewing a promo-clip timeline.js or scene, adapting a scene to another aspect ratio, writing a voice-over script, prompting TTS/music/image models for a promo, setting up theme.json, or fixing a promo-clip QA finding.

Recommend which edit style fits a piece of content, with a viral estimate for each, before anything is built. Use when they type /reel-style, ask «کدوم سبک؟» or «این رو چطوری ادیت کنیم», or send new content (a product, a service, a script, a take) without naming a style.