Coding & Development

CPU-only video subtitles: faster-whisper ASR → any-language translate → ASS burn or SRT/ASS export. No GPU. Claude/DSH agent skill with install + verify frame QA.

Time-aligned TTS dub from a target-language SRT: Volcano (豆包) for Chinese, edge-tts for any language, atempo cue fit.

Deterministic CLI executor for video-cooking: wraps yt-dlp, whisperX, ffmpeg into cook subcommands (download/subtitle/dub).

Chinese voiceover clone dubbing: IndexTTS2 speaker-clone + Demucs BGM keep + bi-directional video retime → cooked/<name>.dubbed.mp4 with bilingual burn. Step 3 of video-cooking.

One-encode final localize cut: burn hardsubs (libass) and/or soft-mux, mix TTS dub over original audio bed.

Router skill: one /video-cooking command chains video-download → video-subtitle (optional dub handoff). Part of the three-skill video-cooking set.

Automated video localization, multi-engine ASR, AI translation, FFmpeg release pipeline

Foreign raw → bilingual cooked: whisperX ASR, agent translate, ASS burn, cloud-srt soft tracks, upload.md + cover. Core of the video-cooking chain.

Codex skill for terminology-aware video translation, bilingual ASS/SRT typesetting, hard-sub MP4 render, and measurable pixel/timeline QA.

AI-powered video subtitle & text overlay skill for Claude Code — Shorts/TikTok/Reels animated titles.

yt-dlp best-quality single mp4 download (YouTube n-challenge + login wall). Companion step 1 before video-subtitle / video-dubbing.

OpenClaw skill for video translation and auto-dubbing with TTS alignment

Claude Code skill: fetch X/YouTube/local video → Whisper ASR → style-guided Claude translate → hard or soft subtitle render with frame QA.

Agent skill from carveragents/show-n-tell

Use when the user wants an image or video prompt written, improved, fixed, critiqued, converted to JSON, or turned into a brief — or mentions cinematic prompt, prompt bank, realism fix, \"looks fake / too AI\", camera or lighting or color-grade direction, camera movement, Kling, Seedance, Veo, Sora, Midjourney, image-to-video, start/end frame, shot list, continuity, or a UGC / influencer / talking-head prompt. A prompt-creation engine: turns a rough idea into a directed, production-ready prompt.

**Page:** https://toolmonsters.github.io/gtm-skills-from-youtube/

Video Forger turns plain-English descriptions into production-ready animated MP4s — product demos, pitch decks, social shorts, data visualizations, narrated explainers — via Remotion. Dashboard + sample videos in README.

Claude skills that drive the [Advibly](https://advibly.com) MCP to generate ad creative end to end: on-brand images, UGC-style video ads, talking-actor videos, and carousels.

產出 AI 生圖／生影片的提示詞文字（不生成圖片本身）。透過一次一題的問答收斂需求，再依內建的八欄框架、角色一致性前綴、風格配方與負面清單，輸出可直接貼到 Gemini、Midjourney、千問、即夢等工具的完整提示詞。當使用者說「幫我做一張圖」「寫個提示詞」「做一支商品短影音」「這張圖怎麼生出來的」「我要這種感覺」，或丟出一張參考圖要仿作時觸發。即使使用者沒有說「提示詞」三個字，只要意圖是產出視覺素材、需要有人幫忙把想法寫成指令，就應該觸發。不負責呼叫生圖工具，也不做既有圖片的後製編輯。

Claude Code skill: records a product demo video of a real user flow in your own web app — transcribed from the real UI, recorded to mp4/gif with Playwright + ffmpeg.

Agent skill from ninime09/product-video-skill

Capture an editing-style reference — a screenshot, a frame, a link, a clip someone liked or hated — read it for its system rather than its content, and write what was extracted into that client's EDIT-STYLE.md so the next edit starts closer. Use when the operator drops in an example and says "make it look like this", "I like how they do captions", "this is the vibe", "don't do what they did here", "here's a reference", "use this style", or hands over a screenshot mid-edit. Also use to review wha

<h1 align="center">Claude Skill Hub</h1>

视频/音频转文字 + 核心提炼。USE WHEN 用户提到：转录、转文字、字幕、视频转文字、音频转文字、抓字幕、YouTube转文字、B站转文字、视频笔记、transcript、提炼视频重点、视频总结。