agentleFS
neskuchnyneskuchny/content_factory/.claude/skills/seamless-insert/SKILL.md

seamless-insert

Skill1 starsChanged 2 days agoMITView on GitHub

Seamless AI inserts into real footage — anchor frames A/B, first/last frame, reversing to return to reality, matching fps, color and grain, hiding seams. Use in flows f02, f04, f05, f06 and wherever an AI clip is joined to real footage.

The file

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    No morphing. A straight cut or a match cut is more honest and looks more expensive.

    line 33

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “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

Seamless AI insert into real video

Anchor frames

  1. Frame A — an exact frame from the source: ffmpeg -ss <T> -i src.mp4 -frames:v 1 -q:v 1 A.png (T comes from the skeleton; prefer a moment without camera or person movement).
  2. Frame B = image_edit(A): only the target zone changes. Check the A/B difference outside the zone — it must match. If it does not, redo the edit.
  3. Static before video: the reviewer checks B before any video generation.
  4. Video A→B: Kling 3.0 Omni (Frames) or Kling 4.0 keyframes. The camera is static in the prompt.
  5. Return: reverse the clip (ffmpeg -vf reverse -af areverse; the audio is replaced anyway). The last frame of the reversed clip = A = a real frame, so the seam is clean.

Fitting the AI clip to the source

# fps and size as in the source, trim the unstable edges
ffmpeg -i ai.mp4 -vf "fps=30,scale=1080:1920:flags=lanczos,trim=start_frame=4,setpts=PTS-STARTPTS" ai_fit.mp4
# common grain over the whole video (after joining)
ffmpeg -i joined.mp4 -vf "noise=alls=6:allf=t" out.mp4
  • Phone HDR (HLG) → convert to SDR first, otherwise the color will "jump".
  • Match the AI clip's color to frame A: eq/curves by hand from 2–3 measurements, or color-matcher. Then apply a common LUT to the whole video.

Hiding the seam

  • Speed ramp of 5–8 frames + a whoosh at the join.
  • Whip pan (motion blur), a flash, or a wipe on a word.
  • Cut on an action or a word ("right here…").
  • No morphing. A straight cut or a match cut is more honest and looks more expensive.

Seam check

The reviewer watches ±10 frames around every join (ffmpeg -ss <t-0.3> -t 0.6 -vf fps=30 seam_%02d.png): a jump in color, brightness, geometry, the person's pose, sharpness.

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.