Edit Videos with Google Omni

Step-by-step workflow

Use Claude + Gemini Omni to recreate any video style with your own script. No editing skills needed.

Tools: Claude AI · Gemini Omni · CapCut · Instagram / YouTube

---

1. Find your reference video

Instagram, Pinterest, or YouTube

Find 1–2 motion design videos whose style you love. Screen-record or save them. Also have your script ready — the exact words you want spoken in your video.

Where to look

Search "kinetic typography reel", "motion design short", or "AI video edit style" on Instagram Reels or YouTube. Save anything that has the energy you want.

---

2. Open Claude and paste this prompt

Upload your reference video (or screenshots from it) and paste your script. Then paste the prompt below. Claude will analyze the style and write ready-made scene prompts for you.

I'm attaching a reference video and my script below.
I want to recreate this video's STYLE for my own
script using an AI video generator (Gemini Omni / Veo).

Step 1: Analyze the reference video deeply. Describe:
- Animation style (2D, 3D, kinetic typography, flat design)
- How things move (speed, easing, bounce, smooth or snappy)
- Transitions (cuts, wipes, morphs, zooms)
- Color palette (exact colors and background)
- Typography style (bold, thin, animated, where text appears)
- Camera behavior (static, zoom, pan, shake)
- Overall mood and pacing (fast, calm, dramatic)
- Sound feel (whooshes, clicks, bass hits, music style)

Step 2: Break my script into scenes of 8-10 seconds
each (because the AI generates max 10 seconds per clip).

Step 3: For EACH scene, write one detailed prompt that includes:
- What appears on screen and how it animates
- The exact text shown on screen (max 5-6 words)
- Colors, background, and typography matching the reference
- Camera movement
- Transition into the next scene
- Sound effects and music mood

Here is my script:
[PASTE YOUR SCRIPT HERE]

---

3. Get your scene prompts from Claude

Claude breaks your script into individual scenes and writes a detailed prompt for each one. Each scene is 8–10 seconds — the max Gemini Omni generates per clip.

Scene	Purpose	Timing	Style notes
Scene 1	Hook / Threat	0 – 8 sec	Bold text · Bass hits
Scene 2	Proof / Reveal	8 – 17 sec	Glitch FX · Zoom punch
Scene 3	Breakdown	17 – 25 sec	Stickers · Paper slap
Scene 4	CTA	25 – 35 sec	Spin FX · Crowd cheer

Copy each scene prompt — they're ready to paste into Gemini.

---

4. Generate in Gemini Omni

One scene at a time

Open the Gemini app. Select the video model — Gemini Omni Flash is the newest. Paste one scene prompt, attach your reference video, and hit generate. Repeat for each scene.

Pro tip: Always attach your reference video or raw footage clip to each generation. Without it, Omni invents the style from scratch instead of matching yours.

---

5. Iterate with chat commands

If a scene looks wrong, tell Gemini in plain language. It will redo that scene. Keep the best version of each clip.

Example fix commands

• "Make the text bigger"
• "Slow down the animation"
• "Change the background color to red"
• "Add more glitch effect on the transition"
• "Make the text slam in harder"

---

6. Assemble in CapCut

Drop all your scenes into CapCut (or any editor), arrange them in order, add music or voiceover, and export. Done.

Important: If any on-screen text looks messy or wrong, redo that text directly in CapCut or Figma. AI still struggles with perfectly clean text — always double-check before posting.