Create Viral Shinkai-Style Anime Videos with ChatGPT, Google Flow, and CapCut
Workflow Overview
Learn how to produce breathtaking 2D Makoto Shinkai and Kyoto Animation-style cinematic anime shorts using ChatGPT, Google Flow, and CapCut. This guide details how to execute a specialized Master Prompt to generate consistent character concepts, car reveals, and lush environments, animating them seamlessly without 3D CGI drift.
🛠️ Tools & Resources Used
- ChatGPT (AI Prompt Director) - Generates aesthetic character concepts, luxury car pairings, and consistent 6-shot anime video prompts.
- Google Flow (Omni Flash) (AI Video Generator) - Renders high-quality 2D cel-animated video clips with authentic hand-drawn anime lighting and line work.
- CapCut (Video Editor) - Assembles video sequences, aligns cinematic beat cuts with ambient audio, and exports the final video in 4K.
⏱️ Quick Workflow Summary
- Step 1: Run the Anime Master Prompt in ChatGPT to generate and select a cinematic concept bundle.
- Step 2: Generate the 6-shot video prompt sequence in ChatGPT with strict 2D visual style enforcement.
- Step 3: Configure Google Flow video settings using the Omni Flash model.
- Step 4: Batch-render Shots 1 through 6 sequentially in Google Flow.
- Step 5: Assemble clips, synchronize beat drops, and mix ambient ASMR audio in CapCut.
đź“ť Step-by-Step Tutorial
Step 1: Pitch Aesthetic Concepts in ChatGPT [Timestamp: 00:00:48]
Open ChatGPT, paste the Master Prompt, and send it. In Phase 1, ChatGPT acts as an expert anime video prompt director and generates four distinct aesthetic bundles detailing the lead male and female characters, signature classic/luxury cars, scenic estate backdrops, and golden hour lighting. Review the options and select your preferred concept bundle to proceed.
Prompt / Settings:
You are an expert AI video prompt director specializing in classic 2D Japanese anime cinematics (Kyoto Animation and Makoto Shinkai style), optimized for Google Flow video generation.
Your task has TWO PHASES:
--------------------------------------------------
### PHASE 1: CONCEPT PITCHING
Provide 4 distinct aesthetic concept bundles. Each set must include:
1. Lead Male & Lead Female (features, expressions, signature outfits in anime aesthetic)
2. Signature Car (vintage, classic, or exotic sports car with exact color/finish)
3. Estate & Location (architecture, courtyard, lush gardens, scenic driveway)
4. Mood & Lighting (golden hour, soft morning sun, blooming petals, lens flare)
Stop here and wait for me to choose one concept (or request edits) before proceeding.
--------------------------------------------------
### PHASE 2: GOOGLE FLOW VIDEO PROMPTS
Once I choose a concept, generate a 6-shot sequence in vertical format (9:16).
CRITICAL STYLE ENFORCEMENT:
The video engine must NEVER drift into photorealism, live-action, or 3D CGI. Every generated prompt MUST strictly embed these anchors:
- Positive Style Anchor:
"Authentic 2D Japanese anime aesthetic, traditional hand-drawn cel animation, Studio Kyoto Animation and Makoto Shinkai style, crisp clean anime line art, vibrant cel-shaded color palette, hand-painted scenic backgrounds, anime keyframe, smooth subtle motion, warm volumetric sun flare."
- Hard Exclusions / Negative Prompts:
"No photorealism, no 3D render, no CGI, no realistic human skin texture, no live-action footage, no Unreal Engine, no Octane Render."
REQUIRED 6-SHOT BREAKDOWN:
- Shot 1 (Macro Detail): Extreme close-up on a luxury anime accessory (watch, cufflinks, smoothing tie).
- Shot 2 (Stepping Out): Anime male lead opening large double villa doors, stepping into radiant garden sunlight.
- Shot 3 (Car Showcase): Low cinematic sweep over the glossy hood, wheels, and badge of the parked car.
- Shot 4 (Driving Sequence): Driver shifting gears, smooth profile view driving down a tree-canopied road.
- Shot 5 (Arrival): Car arriving at an estate courtyard fountain; low angle of dress shoes stepping out.
- Shot 6 (The Meet): Slow-motion reveal through falling petals to the female lead smiling warmly, closing with an intimate romantic profile framing.
Format each shot with:
- Shot Title & Duration
- Direct Google Flow Prompt
- Negative Prompt
Step 2: Generate the 6-Shot Storyboard Prompts in ChatGPT [Timestamp: 00:01:04]
Reply with the concept number you selected. ChatGPT immediately executes Phase 2, breaking down the story into 6 chronological vertical shots: Macro Detail, Stepping Out, Car Showcase, Driving Sequence, Arrival, and The Meet. Each prompt embeds hard negative exclusions and positive Kyoto Animation/Makoto Shinkai style anchors to ensure the model produces authentic cel animation rather than 3D CGI.
Step 3: Configure Video Settings in Google Flow [Timestamp: 00:01:39]
Open Google Flow and create a new project. Switch the mode to Video and configure your generation canvas. Select the Omni Flash model, which excels at clean 2D line work and smooth animation without melting anime character faces. Set the aspect ratio to 9:16, the duration to 10 seconds, the resolution to 720p, and set the generation count to 1.
Prompt / Settings:
Mode: Video
Model: Omni Flash
Aspect Ratio: 9:16 (Vertical)
Duration: 10 seconds
Resolution: 720p
Output Count: 1
Step 4: Render All 6 Shots Sequentially in Google Flow [Timestamp: 00:01:54]
Copy the prompt for Shot 1 from ChatGPT, paste it into Google Flow, and hit generate to render the luxury accessory macro clip. Continue this process for Shots 2 through 6 in sequential order. Because the master prompts embed identical character traits, color palettes, and lighting descriptions, each generated scene maintains perfect visual continuity.
Step 5: Timeline Assembly, Beat-Syncing, and Sound Design in CapCut [Timestamp: 00:02:15]
Import all six rendered clips into CapCut Desktop and arrange them in chronological narrative order on the timeline. Import a nostalgic piano or lofi hip-hop track, then trim your video transitions so visual impacts—such as the car door closing, the engine starting, and footsteps on the courtyard gravel—snap directly to the musical beat drops. Add ambient sound effects (soft wind and rustling leaves) set to -20 dB, then export the final video in 4K resolution.
Prompt / Settings:
Timeline Order: Shot 1 (Macro) -> Shot 2 (Doors) -> Shot 3 (Car) -> Shot 4 (Drive) -> Shot 5 (Arrival) -> Shot 6 (Meet)
Background Music: Nostalgic piano / lofi beat
Ambient Audio Layer: Gentle breeze and rustling cherry blossom leaves (-20 dB)
Export Resolution: 4K (3840x2160 or 2160x3840) / 60 FPS
đź’ˇ Key Takeaways & Pro Tips
- Prevent 3D CGI Drift: Modern AI video models tend to default toward realistic human skin and 3D textures. Always enforce the negative prompt anchor (
No photorealism, no 3D render, no CGI, no realistic human skin texture) across every generation. - Cut on the Beat: The secret to viral anime edits is rhythm. Align physical contact actions (gear shifts, doors shutting, foot steps) with the snare or downbeat of your piano soundtrack to trigger maximum viewer engagement.
- Strict Multi-Shot Narrative Structure: Following the 6-shot progression (from accessory macro detail to romantic meet) gives short-form videos a complete cinematic arc that boosts average view duration.
đź“Ś Original Source & Attribution
- Source Video: How to Make Viral Shinkai-Style Anime Videos ( Step-By-Step )
- Original URL: https://www.youtube.com/watch?v=-LnJFhekPhM&t=48s
- Credit: FlowState AI