agentleFS
alisadiq-aialisadiq-ai/favstash-ai-video-editing-skills/skills/3-editing/edit-video/SKILL.md

edit-video

Skill75 starsChanged 39 days agoApache-2.0View on GitHub

Edit Reels, TikToks and YouTube Shorts from a brief, footage or reference: cut, captions, sound, motion graphics, safe areas and review of the encoded export. The base for every edit; add one style skill (adaptive-glass, paper-grid, breakout-card, product-motion-ads, text-over-footage, animated-explainer, film-styles, 3d-character) for the finished look, and cartoon-character for a lip-synced character.

15 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use cut-mistakes, cut-silences and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • nateherkaiSkill

    cut-mistakes

    Agent 2 of the video editing pipeline. Finds and removes spoken mistakes — stutters, repeated words, false starts, and retakes (re-recorded lines) — from a talking-head recording. Use after cut-silences, when asked to cut mistakes, remove stutters / repeats / filler restarts, clean up flubs, or keep the best take. Works review-gated: it proposes every cut with context and a reason for approval, then renders only the approved cuts via ffmpeg. Requires a word-level transcript.

    Also in 5 projects that use this

  • nateherkaiSkill

    cut-silences

    Agent 1 of the video editing pipeline. Removes silences and dead air from a talking-head recording. Use when asked to cut silences, trim pauses, remove dead air / gaps, or tighten the pacing of a raw video, given a word-level transcript. Produces an edit list (EDL), a re-timed transcript for downstream agents, and optionally the cut video via ffmpeg. Does NOT cut mistakes, repeats, or false starts — that is the cut-mistakes agent.

    Also in 5 projects that use this

  • NexumAI-onlineSkill

    short-form-edit

    Convertí una grabación hablada en un reel, YouTube Short o anuncio corto terminado, con apertura que genere curiosidad, resolución respaldada por el contenido, cortes narrativos, gráficos sincronizados, B-roll en movimiento, subtítulos, música sutil y efectos. Para videos largos usá edit-video; para quitar sólo pausas, cut-silences.

    Also in 4 projects that use this

  • Google CloudSkill

    video-storytelling

    Cinematic video ad storytelling expertise. Covers 3-act narrative structure, emotional arcs, camera motion, scene transitions, and voiceover pacing for 10-second commercial video ads using Gemini Omni. Load this skill when generating video ads or storylines.

    Also in 5 projects that use this

  • ConardLiSkill

    motion-showreel

    用代码制作图形、文字或界面动画视频,输出 MP4 和可编辑源码,最长 20 分钟。适用于动效作品集、产品片、UI 动效、动态排版、Logo 片头、生成艺术与讲解视频;支持独立选择 BGM/音效来源、用户口播,以及用用户提供的口播稿通过 TTS 合成语音。不用于实拍制作、写实人物生成或一般视频剪辑。

    Also in 4 projects that use this

  • chriswesttSkill

    make-a-video

    Beginner-friendly end-to-end video creator for HyperFrames. Use when the user says "make a video", "create a video", "new video", "build a video", "video from scratch", "I want to make a video", "help me create a video", or when someone who's never used HyperFrames before arrives with a concept, script, or rough idea and wants a finished MP4. Interviews the user in one pass, then builds the full video with mandatory preview and visual-verification gates.

    Also in 4 projects that use this

  • +2 more from nateherkai

The file

Show its rules and text

We split this skill into the rules it gives an agent

15 rules. 0 are common, 14 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 14 edited versions we read keep these.

  • 14

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use the creator's own brand and references when they have them.

    line 27
  2. Very specific:

    Read the brief, supplied references and any creator preferences.

    line 32
  3. Distinctive:

    Use the existing editor and run structure when they work.

    line 43
  4. Distinctive:

    Use the creator's brand and references to choose type, color, energy and composition, then the style that fits (above).

    line 47
  5. Distinctive:

    Don't force a split, panels or an editing app on material that doesn't need it.

    line 49
  6. Distinctive:

    Use a reference for timing, composition and sound patterns.

    line 69
  7. Distinctive:

    Check source movement: a speaker stepping back can cancel a digital zoom.

    line 76
  8. Distinctive:

    Default to 1080×1920, 9:16, normally 30 fps unless the brief or source calls for something else.

    line 81
  9. Distinctive:

    Use one coherent type hierarchy and colors that work with the real footage.

    line 101
  10. Distinctive:

    Start captions around 46–68 px on a 1080-wide export, usually one or two meaningful lines, then inspect at phone size.

    line 102
  11. Distinctive:

    Keep production notes in the report.

    line 109
  12. Distinctive:

    Use owned or licensed music.

    line 120
  13. Distinctive:

    When sounds follow on-screen events, put each one on the first frame that shows its event: struck sounds on their attack, swells on their …

    line 141
  14. Distinctive:

    Keep editable source and separate audio/caption tracks when the editor supports them.

    line 147
  15. Distinctive:

    Keep numbered exports and the recoverable project/source.

    line 163

“Common” is measured: kept by at least 6 in 10 of the 14 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Edit video

The base guide for every short-form edit. It owns the cut, captions, sound, motion graphics, safe placement and review. A style skill on top decides the finished look. Make the treatment fit the story and actual footage: a natural reflection can stay simple; a product claim may need moving proof.

Pick a style

StyleLooks likeFits
adaptive-glassSpeaker with moving proof in footage-matched glass panelsProduct, tool and explainer reels with real screen proof
paper-gridWhite grid pane above the speaker, one animated UI artifact per lineTool, repo and feature explainers where every sentence names a thing
breakout-cardGrid page, speaker in a bottom card with their head breaking out, one-to-two-word captionsTalking-head explainers that alternate graphics and personal lines
product-motion-adsMotion graphics built from the product itself, one continuous piece where each scene turns into the nextPromoting something the creator built, free resources and comment-to-DM giveaways; no camera footage needed
text-over-footageOne readable line over B-roll that matches itJokes, observations and short setup/payoff reels
animated-explainerFully animated: a lit colour field and one big hero card per spoken beat, the line word-lit at the top, keyed to a locked voiceTool, product and idea explainers and two-voice conversations when the creator doesn't film
film-stylesOne of 99 film looks drawn in code: whiteboard, stick figures, tablet sketch, chalkboard, watercolour, cel anime, manga, newsreel, caption news, data stories, 3D maps, isometric, tilt-shift, brick toy, risograph, neon and moreA whole reel, or a 16:9 YouTube explainer in chapters, with no camera; each style says what goes first
3d-characterA cute 3D chibi kid (boy or girl) talking to camera on a pastel stage, camera cuts on the words; Three.js, no BlenderPlayful explainers and announcements when the creator doesn't film
cartoon-characterA lip-synced 2D cartoon that talks, blinks and gestures (the creator's own, made with AI, or a sample)A cursor in animated-explainer, a corner presenter on any reel, or a cartoon explainer

Use the creator's own brand and references when they have them. Without a style installed, follow this guide alone with clean full-camera framing.

Start from what exists

  • Read the brief, supplied references and any creator preferences. A brief from $write-script carries the approved openings, body, proof map and must-keep claims. Resume the existing project for revisions and preserve prior cuts. Keep approved copy and the mix during a layout-only revision.
  • Inspect representative frames and listen to the takes when playback is available. Transcribe speech when needed; preserve meaning, qualifiers and natural pauses. Do not invent spoken words or silently rewrite an approved claim.
  • Find retakes before cutting. Creators often restart a line and say it again in full; keep the later complete attempt. One transcript of a whole take can hide a restart, so split at pauses and transcribe the chunks separately.
  • Use the existing editor and run structure when they work. For a fresh project, keep inputs, assets, editable source, previews, exports and review notes in one dated folder. The local helpers can create this; they are optional, and editing does not require an onboarding questionnaire.
  • Use the creator's brand and references to choose type, color, energy and composition, then the style that fits (above). Keep personal reflections simple. Don't force a split, panels or an editing app on material that doesn't need it.

Build the cut

Find the strongest honest opening and the minimum context needed for its payoff. Choose each shot for what the viewer needs to see at that moment:

MaterialUseful treatment
Personal delivery or emotionFull camera; remove false starts while retaining meaningful breaths, humor and pauses
Voice plus a visible action/resultSpeaker and proof together while both remain legible; expand the proof for detail
Desktop or phone demoCrop to the actual action, preserve spatial continuity and hold the result long enough to inspect
Copy over footageMatch action or emotion to the line; give each phrase time to read without adding an unnecessary speaker
Relationship or idea that needs explanationA purposeful animated insert; see motion graphics

Own the supporting work: source, capture, trim and inspect the footage available within the task. Ask for a specific missing asset only when access or recording requires the creator. Show real product behavior; label mockups and conceptual graphics. Check screen captures for private data across moving frames.

Use a reference for timing, composition and sound patterns. A transcript alone does not establish its visual treatment. See reference media when analyzing a URL or reusing third-party material. Keep source/rights notes for output assets; downloadable does not mean reusable.

Design the opening's picture, emphasis and sound together. For a tension/reveal hook, make the initial push visibly register and resolve the build at the payoff. Check source movement: a speaker stepping back can cancel a digital zoom. A framing grid or kinetic phrase can suit the reference, but is not a universal style.

Compose for a phone

Default to 1080×1920, 9:16, normally 30 fps unless the brief or source calls for something else. Keep faces proportionate and complete; inspect crops and blurred fills for a second ghost face. Enlarge useful detail instead of fitting an unreadable desktop into a tiny panel. Screen proof normally fills its pane; decorative cards and headings should not shrink or obscure it. Use full-screen proof when detail needs it. Change layout on meaningful beats.

For an unknown destination, use this conservative working mask at 1080×1920: essential text, captions and demonstrated UI detail within x=72..900, y=269..1248. Scale proportionally for another 9:16 resolution. This is a house guide, not an official platform specification; check the intended app's current preview before publication. Footage and decorative panels can extend beyond it.

Center main captions and cards on the whole canvas, x=540. A centered critical box can be at most 720 px wide under that mask. Inset or wrap essential contents inside wider panels instead of shifting the whole design left. Measure the whole text box, backing, outline and shadow, including animated extremes; a safe anchor alone does not establish safe placement. Keep eyes, mouth and focal actions clear of captions and expected app controls.

Use one coherent type hierarchy and colors that work with the real footage. Start captions around 46–68 px on a 1080-wide export, usually one or two meaningful lines, then inspect at phone size. Preserve the actual words and keep caption changes aligned with speech. See captions for detailed timing, translations or caption revisions. Avoid competing transcript layers. Alongside captions, normally use only one brief supplementary emphasis. Headings, eyebrows and footers are optional, not a template to fill. Demonstrate the feature through motion or actual caption behavior instead of explaining the edit in text. Keep production notes in the report; disclose a mockup on screen when needed to prevent it being mistaken for real evidence. When B-roll is absent, animate the relevant idea rather than leave an empty pane or substitute a text-heavy slide. Simplifying does not mean one oversized icon: show connected state changes with relevant assets, such as footage entering a timeline and gaps closing. Carry an object through related beats; keep compact object labels subordinate to speech.

Sound and motion

Speech comes first. Trim with short fades, preserve natural tone and remove unwanted source/demo audio. Duck music and SFX under speech; silence is valid. Use owned or licensed music. If the creator will add native platform music, deliver the appropriate cut without a music bed and note the intended timing.

Make SFX follow visible events: click for a UI action, pop for an element arrival, shutter for a meaningful camera/cut change, swish for a motivated zoom, tick for a step, thock when something settles, latch when it locks or connects. A short riser can build into a reveal; resolve it at the visual payoff. Merge nearby cues and keep ordinary caption changes silent. Prefer the creator's established sounds; otherwise start with the SFX pack. Copy only selected files, align audible transients rather than file starts, and audition them against speech. The pack's other one-shots (ding, sparkle, impact, typing tick, ratchet, shimmer) cover payoffs, typing and product moments. Suggested gains are starting points, not a volume ceiling: a demonstrated click or reveal build must be audible. Trim/time-stretch a riser to the reveal and check the combined mix, not isolated cue levels.

Animated reels (animated-explainer, film-styles, 3d-character) start from a locked voice: scripts/voice.py places one take per line and writes every word's time, from the creator's own recording, a free local voice or ElevenLabs. See voice options, including how to clone your own voice for free.

When sounds follow on-screen events, put each one on the first frame that shows its event: struck sounds on their attack, swells on their peak. Sound on the frame explains the rules, and scripts/sfx-sync.mjs measures sync points, mixes a cue sheet under the voice, masters it with encoder headroom and draws strips that prove every hit on the export.

Keep editable source and separate audio/caption tracks when the editor supports them. Use FFmpeg for straightforward trims/assembly and HyperFrames when a coded composition helps; each style skill brings its own builder. Read local tools when setup, analysis or rendering helpers are needed, including its render reliability notes. Render a representative treated sample, inspect it, then finish the requested cut; do not multiply variants before it works.

Review and deliver

Inspect the encoded candidate, following review: frame 1 and the first 3 seconds of every variant, changing layouts, longest captions, moving proof, sound and ending. Verify streams and full decode as well as visual quality. Report only checks actually performed; measurements do not prove a listening pass.

Keep numbered exports and the recoverable project/source. Show the actual video, summarize what changed and state remaining limitations. Technical validity, creative acceptance and publication approval are separate.

Editing works without FavStash. For an approved cut, publish-and-analyze owns the delivery export, covers, variants, scheduling and results. Never publish, schedule or upload merely because a cut is finished or a service is connected; obtain approval for the exact outward action.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.