Agent skill · creative production · samuraigpt
muapi-freeze-effect-video
Generate a cinematic "freeze effect" video where time stops mid-scene, the subject walks through the frozen world, then time resumes with a snap.
Why this skill is useful
Adds specific video generation commands and creative direction for a unique cinematic effect that the AI wouldn't reliably generate on its own.
What it needs
Requires ffmpeg installed locally. About 3k tokens when loaded. Last updated 2026-08-04. 3,990 stars on the source repository.
What this skill does
Freeze Effect Video Generate a cinematic "freeze effect" video where time stops mid-scene, the subject walks through the frozen world, then time resumes with a snap. Inputs Name Type Required Default Description :--- :--- :--- :--- :--- photo imageurl yes — Reference photo of the main character (face/body) whose appearance must remain consistent across all shots. scene text no a packed, dimly lit sports bar with neon accents, blurred TVs showing a championship celebration The setting where time will freeze (e.g. "a busy nightclub", "a wedding reception", "a stadium crowd"). freezemoment text no golden arcs of beer suspended midair, popcorn floating motionless, people frozen mid-cheer The signature frozen visual — what hangs in the air when time stops. closingline text no perfect A short word/phrase the character whispers before resuming time. aspectratio text no 16:9 Output aspect ratio — "16:9" for cinematic, "9:16" for vertical/social. duration text no 15 Target duration in seconds (10 or 15 recommended). Steps Phase A — Generate the Freeze Effect Video Submit the plan with ONE step: 1. Freeze effect video — muapi video from-image (model=seedance-v2.0-i2v, quality=high, generateaudio=true): Reference image: {{photo}} Aspect ratio: {{aspectratio}} Duration: {{duration}} Resolution: 720p Prompt: After generation, present the video to the user. Notes The character's identity is anchored by @image1 — the reference photo must clearly show the main subject's face. For best results, choose a scene with crowd energy and lots of small objects (liquid, confetti, sparks) that can plausibly "freeze" in midair. If the model rejects realistic human likeness, switch to the Global tier model (seedance-2-image-to-video-fast) which allows looser identity matching. For vertical/social delivery, set aspectratio=9:16 and keep duration=10 for tighter pacing. Always keep generateaudio=true — the snap → silence → snap audio arc is the signature of this effect and should not be muted. …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills samuraigpt/freeze-effect-video