Video & audio

Video editing

Cutting judgment for short-form: pacing, B-roll, colour, keyframes, export settings.

Paid $29 list, $23 until 2026-12-31 · one skill, yours for good. The outline, the steps and the file open after purchase.

Background

Tuned for the three mainstream platforms Xiaohongshu (RED) / Douyin / WeChat Channels. An end-to-end pipeline source ingest → voiceover → story restructure → platform gates → auto enrichment → render → three-platform export → title and body copy, with parameters tuned to each platform's algorithm, aspect ratio, duration and review rules — not just an editing tool.

Why it is built this way

The pipeline is long because each step leaves a receipt a later step can re-check. Every script follows one shape, plan then apply then confirm then verify --strict, and verification re-reads the source hash, decodes the whole file again and compares media contracts, so a run that stopped halfway cannot pass for a finished one. production_authorization binds scope, assets and rights to SHA-256 digests before anything reaches an external service, and changing a byte or a provider invalidates the old report.

The render sits at the very end as one encode on purpose. The older route split the video, burned subtitles, merged clips, then made a cover and a chapter bar, which is four or five re-encodes and a visibly soft final cut, so subtitles, overlays, focus events, picture-in-picture and loudness normalisation became flags on render_final.py instead. Whatever the gates cannot judge stays with a human: ready_for_human_review only claims the technical checks passed, and platform gates run before the render.

Where it bites

• Mixed frame rates joined with -c copy A 30 fps drone clip next to a 29.97 fps phone MOV, joined with -c copy, freezes the picture around the one-minute mark because timestamps are not continuous across the seam. clip_assembly.py plans a single encode that normalises canvas, CFR, SAR and audio layout, and the seam proof has to be watched before confirm.
• Audio that stops mid-film Combining amix silence padding, atempo and video processing in one filter_complex can end the audio at 1:55 while the picture keeps playing. The document splits the chain: adelay for the cover segment, atempo for the speed change, apad to the target duration, export a WAV, then mux with -c:v copy.
• End card cut off by -t B-roll assembled to 431 seconds for a 215-second voiceover puts the end card past the -t limit, so it never shows. The fix is to trim the B-roll to the measured audio duration first, then join cover, trimmed B-roll and end card, with every segment's duration known before the join.
• A word split inside a subtitle Chinese wrapping counts characters, so ASCII letters get the width of a hanzi and OpenClaw breaks mid-word at the line boundary. The wrapper needs display width, one unit per CJK character and about half per ASCII letter, plus a word-boundary check that refuses to break inside a run of letters.

When to use it

Xiaohongshu/RED-tuned short-form video workflow for voice-over, talking-head, tutorials, explainer and science-popularisation videos assembled from external footage (official product demos, screen recordings) as well as self-shot material, interviews, podcasts, screen recordings, chroma key, B-roll, captions, and generated assets. Covers source ingest and conformance, hard and ping-pong loops, edit routing and runtime profiles, transcription and semantic review, multi-take and multicam sync/speaker switching, stabilization/dead-air cleanup, highlights/shorts, story/source/authorization gates, B-roll and generated-asset planning/review, storyboard animatics and shot handoffs, audio design/mix/loudness, captions, color/speed/J-L cuts, local upscale/interpolation review, reversible revisions, render/QA, multi-platform and size exports, publish packages, NLE handoff, and Remotion.