Generate covers, vertical recomposes, SEO packs, and chapter cards from edited videos.
npx skills add https://github.com/openakita/openakita --skill media-post
Please help me install this Agent Skill. Skill: media-post Source: https://github.com/openakita/openakita/tree/main/plugins/media-post Command: npx skills add https://github.com/openakita/openakita --skill media-post
Turning a finished video into platform-ready publishing material requires manual cover selection, aspect-ratio re-editing, copywriting for each social platform, and chapter card design. This Skill automates that entire post-edit packaging workflow through four modes backed by Qwen-VL-max vision scoring and ffmpeg.
Ask the assistant to use media_post_create with mode cover_pick on your edited video file to get scored cover candidates.
High-intent search queries and answers about installing and using this skill.
Use the cover_pick mode, which prefilters about 30 candidate frames with the ffmpeg thumbnail filter, then scores each on six aesthetic axes via Qwen-VL-max. The top-N frames above your min_score_threshold are copied to the task output folder.
The multi_aspect mode detects scene cuts with ffmpeg, tracks the main subject using Qwen-VL-max bounding boxes, smooths the crop trajectory with EMA, and renders a dynamic crop. It currently outputs 9:16 and 1:1 only.
Only the chapter_cards mode works without an API key since it renders locally. cover_pick, multi_aspect, and seo_pack all require a DashScope key configured in the plugin Settings tab.
The renderer transparently falls back to ffmpeg drawtext, so cards still generate without Playwright or CJK fonts. The output is visually blockier but valid, and the render_path field records which path was used.
A 30-second clip at fps=2 costs about 0.32 CNY, while a 30-minute clip costs roughly 35 CNY. Estimates above the warn threshold of 10 CNY require explicit cost_approved=true before the task runs.
It only emits 9:16 and 1:1 ratios, recommends source videos of 30 minutes or less, and caps ffmpeg crop expression nesting at 95 levels with automatic downsampling for long segments. Ratios like 3:4 and 21:9 are deferred to a later version.
KaiserWhoLearnsConvert media files between containers and codecs using FFmpeg.
luzhenqianGenerate episode cover thumbnails and per-platform metadata from a single preset.
NoizAITranslate and dub spoken audio in videos using subtitle files and timeline-based TTS.
anavvanzinCoach short-video editors through post-production workflows in CapCut Pro, Premiere Pro, DaVinci Resolve, and Final Cut Pro.
