_____                   _____            .__
  /     \ _____  ___  ____/ ____\_ __  _____|__| ____   ____
 /  \ /  \\__  \ \  \/  /\   __\  |  \/  ___/  |/  _ \ /    \
/    Y    \/ __ \_>    <  |  | |  |  /\___ \|  (  <_> )   |  \
\____|__  (____  /__/\_ \ |__| |____//____  >__|\____/|___|  /
        \/     \/      \/                 \/               \/

Ad Restyler

A Claude skill that restyles any ad or video into a completely different visual world — Aardman claymation, LEGO-style bricks, X-ray, 16-bit pixel art, Ghibli anime, felt puppets, voxel art, knitted yarn, premium Pixar-style 3D — while keeping every shot, camera move, performance, and beat of motion exactly as filmed.

You give it a video and pick a style. The skill walks a fixed gate flow — source check, style selection, preservation preferences, prompt approval — then either generates the restyled clips itself through the MaxFusion MCP (Google Omni Flash, video-to-video) and returns them in chat, or hands you the complete run-ready prompt package for your own Google AI Studio key. Either way you end the conversation with everything you need.

Examples

Original	Restyled
source ad	Aardman Claymation
source ad	Felt Puppet Stop-Motion
source ad	Plastic Construction Bricks (LEGO-style)

Full-quality MP4s are in examples/.

What the skill enforces

• The Core Formula — every restyle prompt follows one locked skeleton: style name, material and rendering details, and a mandatory verbatim invariance clause ("Keep everything else exactly the same — same shots, same camera moves, same motion."). That clause is what makes this a restyle instead of a regeneration.
• A five-gate flow — source check, style selection, preservation preferences, prompt approval, delivery choice. Nothing generates before you approve the exact prompt.
• Materials only, never content — style details describe surfaces, textures, and rendering. No new actions, characters, objects, settings, or text ever enter a prompt.
• Performance preservation — when the source has people whose facial performance and comic timing matter, a dedicated preservation line is appended so the acting survives the restyle.
• Segment consistency — ads longer than 10 seconds are split into segments, and every segment runs the identical prompt. Consistency comes from prompt identity; the style wording never varies between clips.
• One style per prompt — never blended, unless you explicitly ask for a hybrid custom style.
• A 9-preset library + custom style builder — styles not in the library are built live through the same Core Formula and shown to you before anything runs.
• MCP discipline — on the MaxFusion path: live capability check before generating, 720p locked, clips delivered as inline widgets in chat.

Requirements

• Claude Code (or any agent that can follow the skill file).
• MaxFusion MCP — video generation runs through it on the default path.

Install

1. Drop SKILL.md into your skills directory:

mkdir -p ~/.claude/skills/ad-restyler
curl -o ~/.claude/skills/ad-restyler/SKILL.md \
  https://raw.githubusercontent.com/holy-templar/ad-restyler/main/SKILL.md

2. Connect the MaxFusion MCP server:

claude mcp add --transport http maxfusion https://mcp.maxfusion.ai/mcp

The first call will ask you to authenticate with your MaxFusion account.

3. In Claude Code, say something like:

restyle this ad as claymation — and attach your video.

The skill takes it from there: style menu, preservation preferences, prompt approval, then generation.

Using it without MaxFusion

The skill file is written for the MaxFusion MCP tool surface, but the pipeline is portable. At the delivery gate, answer no to MaxFusion generation and you'll need:

• a Google AI Studio API key — for the Google Omni Flash video-to-video model that performs the restyle.

If your agent can execute against that API, it runs the generation for you. If not, the skill delivers the complete run-ready package in chat — the final prompt per segment plus run instructions: upload your clip (≤10s), paste the prompt, generate at the highest available resolution. Prompts, gates, and rules stay exactly the same.

License

GNU AGPL v3.0 — Copyright (C) 2026 Maxfusion AI.

Free to use, study, modify and share. If you modify it and let others use it — including over a network — the modified version must be released under the same license.