AI Video Restyle Skill for Talking-Head Videos

Just ask

Say the word, it runs.

The skill fires on natural requests like these. No commands to memorize.

“Restyle this video”“Turn my talking head into anime”“Give this clip a viral podcast edit”“Apply a visual style while keeping the lip-sync”

One clip, eleven styles

Keep the performance. Change the world.

The first tile is the source. Every result beside it keeps the same face, expressions, gestures, speech timing and audio while applying one complete visual system.

Before

Original clip

The same performance, audio and timing used by every style.

Cutout-real

Podcast Pop

Sticker-cutout subject, bold rotating backgrounds, punch-ins and torn caption bands.

Full repaint

Watercolor Wash

A living watercolor portrait with pigment blooms, paper grain and soft painted transitions.

Full repaint

Anime Manga

Crisp cel shading, expressive line art, screentones, speed lines and impact frames.

Cutout-real

Newspaper Print

Black-and-white halftone newsprint, headline columns and sharp red marker accents.

Cutout-real

Notebook Doodle

Ruled paper, hand-drawn ink doodles, highlighter marks and sticky-note energy.

Cutout-real

Neon Vaporwave

Synthwave gradients, a neon grid horizon, chrome type and pink-cyan rim light.

Full repaint

Comic Book

Pop-art ink outlines, Ben-Day halftone dots, starburst backdrops and ZAP caption boxes.

Cutout-real

Vox Style

A signal-yellow editorial field of Bauhaus geometry, photoreal sticker cutout and bold caption bars.

Full repaint

Psychedelic

Liquid marbled rainbow swirls, a posterized portrait and hypnotic color surges.

Full repaint

Illustration

Clean flat vector repaint in an idyllic illustrated scene with drifting clouds and soft skies.

Cutout-real

Cardboard Style

Torn cardboard edges, corrugated kraft, packing tape, stamps and marker doodles.

The pipeline

What the skill does for you.

1
Choose one visual world
Pick Podcast Pop, Watercolor Wash, Anime Manga, Newspaper Print, Notebook Doodle, Neon Vaporwave, Comic Book, Vox Style, Psychedelic, Illustration, or Cardboard Style. One preset governs the whole video so every cut feels designed together.
2
Analyze the performance
The skill reads the transcript, sentence boundaries, emphasis words, source dimensions, duration, faces, marks, and existing captions before it spends on a render.
3
Approve the segment plan
Speech is packed into contiguous 8 to 10 second sections. Each row shows the exact line, caption, look variation, and the words that trigger a punch-in, background swap, impact frame, or type reveal.
4
Restyle beat by beat
Gemini Omni Flash repaints each segment from the source motion. Identity, expression, hairstyle, clothing, hand gestures, props, mouth movement, and lip-sync are anchored in every prompt.
5
Conform every cut
Each generated segment is trimmed or padded to its exact source duration. This removes fractional timing drift before the pieces are assembled.
6
Restore the original audio
The conformed visuals are stitched in order and the untouched source audio runs across the final composition, keeping the delivered video locked end to end.

Six steps, two of them yours.

Pick the look and approve the segment plan. The restyle, the timing fix and the audio lock run themselves.

Start creating

Before and after

The take stays. The world changes.

In the middle is the source clip. On either side, the same ten seconds after the skill. The person, the expressions and the mouth movements carry over; only the visual world is new.

Face, hair, clothing, gestures and props are anchored in every segment prompt
Each restyled segment is trimmed or padded to its exact source length, so lip-sync holds to the end
The untouched source audio runs across the finished cut, never a resynthesized voice
Captions are baked in and checked for spelling, or left off for clean footage

Start creating Install the skill

Illustration

Original

Vox Style

The model stack

Only Google’s Omni models repaint a real take.

Gemini Omni Flash is the default: it repaints each segment from the source motion. Ask for Gemini Omni Flash 1.1 and the agent runs the restyle on the newer model, at a resolution you pick up to 4K. Gemini reads the performance before a credit is spent, ffmpeg conforms the timing, and the Advibly composer lays your original audio back over the cut.

What you get

A finished restyled video with the source performance and lip-sync intact
The exact original audio, restored across the complete cut
Every approved restyled segment saved separately in your library
An editable composition that reopens in the Advibly video editor

Start creating Read the source

Gemini Omni Flash 1.1New

The newer restyle, at the resolution you choose

Gemini Omni Flash

The default: identity-preserving video-to-video restyle

Gemini

Transcript, emphasis and sentence-boundary analysis

Install

Add it in seconds.

Install the skill, then make sure the Advibly MCP is connected so it has tools to call. Works with every skills-capable agent.

Claude Code

Claude

Cursor

Codex

VS Code

OpenClaw

Hermes

ChatGPT

Connect the Advibly MCP first

One click, no terminal

Download .skill Download .zip

Drop the .skill file into Claude → Settings → Skills, or unzip the folder into ~/.claude/skills.

One command (any agent)

npx skills add advibly/skills -s video-restyle

Works with Claude Code, Claude Desktop, Cursor, and any agent that supports skills. Pulls straight from GitHub.

Manual (Claude Code)

git clone https://github.com/advibly/skills.git
cp -R skills/video-restyle ~/.claude/skills/

Or in Claude Desktop: Settings → Skills → Install and select the skill folder.

Bundled references

The playbook.

Each skill ships with reference docs the agent loads on demand, so the craft is baked in.

references/podcast-pop.md

Sticker-cutout subject, bold rotating backgrounds, punch-ins and torn caption bands.

references/watercolor-wash.md

Full watercolor repaint with pigment blooms and living paper texture.

references/anime-manga.md

Cel-shaded repaint with screentones, speed lines and impact frames.

references/newspaper-print.md

Halftone newsprint, editorial columns and red marker accents.

references/notebook-doodle.md

Ruled paper, hand-drawn doodles, highlighter marks and sticky notes.

references/neon-vaporwave.md

Synthwave gradients, grid horizons, chrome type and neon rim light.

references/comic-book.md

Pop-art repaint with ink outlines, Ben-Day dots and starburst panels.

references/vox-style.md

Sticker cutout on signal-yellow Bauhaus editorial layouts and caption bars.

references/psychedelic-swirl.md

Posterized repaint over liquid marbled rainbow swirls and color surges.

references/flat-illustration.md

Flat vector repaint in idyllic illustrated scenes with drifting clouds.

references/cardboard-cutout.md

Torn-edge cutout on corrugated kraft with tape, stamps and marker doodles.

references/pipeline.md

Cutting, upload, billing, conformance, audio-lock, and failure triage.

Honest limits

What to watch for.

Gemini Omni Flash accepts source segments from 4 to 10 seconds, so longer videos are split at sentence boundaries and reassembled.
One preset is used per delivered video. Variation comes from that preset's palette, backdrop, and motion axes.
Recognizable public figures and third-party marks can be blocked by video-to-video moderation.
Model-rendered captions are verified and re-rolled if misspelled. Clean footage with post captions is also supported.
The assembly supports up to 12 scenes per composition, which covers roughly two minutes when segments are packed well.

Questions

Before you install it.

What does the Video Restyle skill do?

It takes a finished talking-head or UGC video and repaints its whole look in one of eleven presets, from Podcast Pop and Anime Manga to Newspaper Print and Cardboard Style. The person, their expressions, their lip-sync and their original audio stay as recorded. Gemini Omni Flash does the repaint, one 8 to 10 second segment at a time.

What do I have to give the agent?

The source video (a local file, a URL, or something already in your Advibly library), the brand to file it under, one preset, and whether you want captions baked in or clean footage. The best source is one continuous shot of someone speaking clearly to camera.

How long can my video be?

Gemini Omni Flash accepts 4 to 10 seconds of source at a time, so the skill cuts at sentence boundaries and reassembles the pieces. One composition holds up to 12 scenes, which covers roughly two minutes of speech. A longer source is split across more than one composition.

What does it cost?

The skill itself is free and open source. Running it spends your existing Advibly credits on the restyled segments, roughly 0.4 credits per second of video, so a 60 second clip runs about 24 credits plus any re-rolls. Every generation shows its credit cost before it runs.

Which agents can run it?

Any agent that supports installable skills and the Model Context Protocol: Claude Code, Claude, Cursor, Codex, VS Code, OpenClaw, Hermes and ChatGPT. Install with one npx command, then connect the Advibly MCP so the skill has tools to call.

Can it restyle any video?

Real people are fine. Recognizable public figures and third-party logos are blocked by the video-to-video content filter, and there is no fallback model, so the skill checks for both while it analyzes your clip and warns you before anything is spent.