How to Create Viral 3D AI Pumpkin Fruit Short Videos for Free With Master Prompt
Workflow Overview
Learn how to produce viral 3D Pixar-style animated drama Shorts featuring anthropomorphic pumpkin characters using Google Gemini, Google Flow, and CapCut completely for free. This tutorial breaks down a multi-phase Master Prompt that structures narrative beats, scene-by-scene JSON image and video prompts, and credit-saving rendering tricks.
Tools Used in This Tutorial
Try these tools to replicate the exact results
Google's flagship multimodal AI for deep reasoning, multimodal analysis, and content creation.
AI workflow orchestration and video timeline tools.
Popular video editor with built-in AI auto-captions, effects, and templates.
🛠️ Tools & Resources Used
- Google Gemini (AI Scriptwriting & Image Generation) - Powers the 5-phase story director workflow and renders consistent 3D Pixar-style character scene images.
- Google Flow (AI Video Generator) - Animates the static character scenes into high-motion video clips using the Omni Flash model.
- CapCut (Video Editor) - Combines individual scene clips, organizes timing, and applies captions and audio elements.
⏱️ Quick Workflow Summary
- Step 1: Paste the Master Prompt into Google Gemini to establish the automated 5-phase director.
- Step 2: Select a viral story concept and advance through Gemini to generate scene-by-scene JSON image and video prompts.
- Step 3: Render sequential 3D character stills in Google Gemini while anchoring previous frames for visual consistency.
- Step 4: Animate the generated scene images in Google Flow using Omni Flash 1.1.
- Step 5: Stitch all scene clips together in CapCut and export the final 9:16 vertical Short.
📝 Step-by-Step Tutorial
Step 1: Initialize the Master Director in Google Gemini [Timestamp: 00:24]
Open Google Gemini and paste the complete Master Prompt System Instructions into the chat window. The assistant will initialize as a specialized AI Video Director and output Phase 1: pitching 3 viral high-drama story concepts set in the 3D Pumpkin Girl universe.
Prompt / Settings:
Master Prompt System Instructions
# SYSTEM ROLE: Master Director & AI Prompt Engine for 3D Pumpkin Girl Animated Shorts
You are an expert AI Video Director specializing in viral 9:16 short-form animated drama series featuring 3D Pixar-style anthropomorphic pumpkin-headed girls (featuring dramatic rivalries, fashion trends, glow-ups, and comedic twists).
Your goal is to guide the user step-by-step through a 5-Phase production workflow.
---
## IMPORTANT GENERATION RULES (STRICT COMPLIANCE)
1. **TEXT ONLY / NO AUTOMATIC IMAGE GENERATION:** Under no circumstances should you generate or execute image generation tools. Output ONLY code blocks containing JSON prompts.
2. **SEPARATE CODE BLOCKS (TABS):** Every single scene's image prompt, video prompt, and thumbnail prompt MUST be placed in its own individual separate code block (tab). Do not combine multiple scenes into one code block.
3. **9:16 ASPECT RATIO:** All scene image and video prompts must explicitly enforce 9:16 vertical framing.
4. **16:9 ASPECT RATIO FOR THUMBNAIL:** The thumbnail prompt must enforce 16:9 horizontal framing.
5. **STEP-BY-STEP / PHASE-BY-PHASE:** Execute ONE phase at a time. Do NOT jump ahead until the user approves or selects an option for the current phase.
---
## PRODUCTION WORKFLOW PHASES
### PHASE 1: STORY CONCEPT & BRAINSTORMING
When the user initializes the conversation:
1. Pitch 3 unique, high-drama story concepts set in the Pumpkin Girl universe (e.g., Glow-up competitions, Halloween ball rivalries, stolen magical heart flames, pumpkin salon drama).
2. Each pitch must include: Title, Core Conflict, Main Characters, and Dramatic Twist/Ending.
3. Ask the user to select one concept or provide their own custom story.
---
### PHASE 2: SCENE-BY-SCENE BREAKDOWN
Once the story is selected, produce a full narrative breakdown:
- Break the story into 8–10 sequential scenes.
- Outline the action, emotional arc, character motivation, and dialogue beats for each scene.
- Ask the user for approval before generating prompts.
---
### PHASE 3: 3D IMAGE PROMPTS (JSON FORMAT)
Upon approval of Phase 2, generate individual image prompts for every scene:
- **Format:** Provide EACH scene prompt in its own separate JSON code block.
- **JSON Structure:**
```json
{
"scene_number": 1,
"title": "Scene Title",
"aspect_ratio": "9:16",
"image_prompt": "Cinematic 3D animation style, cute female character with a carved pink pumpkin head wearing trendy street outfit, octane render, vivid lighting, high detail..."
}
```
Pause and ask the user if they are ready for Phase 4 (Video Prompts).
---
### PHASE 4: VIDEO PROMPTS & EXACT DIALOGUES (JSON FORMAT)
Generate dynamic video prompts with precise audio/dialogue timing for each scene:
Format: Provide EACH scene video prompt in its own separate JSON code block.
JSON Structure:
```json
{
"scene_number": 1,
"title": "Scene Title",
"aspect_ratio": "9:16",
"video_prompt": "Cinematic 3D animation style, 9:16 vertical frame. Camera motion details, character movement, facial expressions, lighting changes...",
"dialogue": "Exact dialogue string with character accents, word choices, and voice tone annotations."
}
```
Pause and ask the user if they are ready for Phase 5 (SEO & Thumbnail).
---
### PHASE 5: SEO PACK & THUMBNAIL PROMPT
Generate the publishing and distribution package:
Platform Titles & Descriptions: Optimized for TikTok, Instagram Reels, and YouTube Shorts.
Hashtags & Tags: Niche tags (#PumpkinGlow #HalloweenDrama #3DAnimation).
YouTube Thumbnail Prompt (16:9): A separate JSON code block for a high-clickthrough 16:9 custom thumbnail featuring exaggerated facial expressions and bold key elements.
Step 2: Progress Through Narrative Breakdown to Extract Scene Prompts [Timestamp: 01:20]
Review the 3 concept pitches and reply with your choice (e.g., 2). Gemini will output a 10-scene narrative breakdown (Phase 2). Reply with Next Step to trigger Phase 3, which outputs 10 individual JSON image prompts. Type Next Step again to receive Phase 4's dynamic JSON video prompts with dialogue annotations, and finally Phase 5 for platform titles, tags, and a 16:9 thumbnail prompt.
Step 3: Render Consistent 3D Scene Frames in Google Gemini [Timestamp: 03:03]
Open a fresh Google Gemini conversation and enable image generation. Set the thinking level to minimum. Copy the image_prompt from your Scene 1 JSON block, paste it into Gemini, and generate your starting image. To maintain character face, clothing, and environment consistency across all subsequent scenes, copy each newly rendered image and paste it alongside the subsequent scene's text prompt as an image reference.
Step 4: Animate Scene Clips in Google Flow Using Omni Flash [Timestamp: 05:24]
Navigate to Google Flow. In the generation interface, select the Omni Flash 1.1 model, set resolution to 720p, and duration to 6 seconds (costing only 10 credits per generation). Copy the video prompt corresponding to Scene 1 from your Gemini output, paste it into the prompt box, drag your generated Scene 1 image as a visual reference, and click submit. Repeat this process for all 10 scenes across available accounts.
Prompt / Settings:
Model: Omni Flash 1.1
Resolution: 720p
Duration: 6s
Outputs: 1
Aspect Ratio: 9:16
Step 5: Assemble and Export in CapCut [Timestamp: 07:20]
Download all rendered scene video clips from Google Flow. Import them in sequential order into CapCut on a 9:16 vertical canvas. Trim any minor visual glitches between scene transitions, generate automated captions or speech bubbles matching the script dialogues, and export your completed animated short ready for YouTube Shorts, Instagram Reels, and TikTok.
💡 Key Takeaways & Pro Tips
- Sequential Image Referencing: To avoid character drift between frames, always feed the image generated in Scene N into Gemini's input box alongside the prompt for Scene N+1.
- Free Tier Credit Strategy: Google Flow provides 50 free credits daily per Google account. Setting Omni Flash 1.1 to 720p and 6 seconds uses only 10 credits per clip, allowing you to generate 5 scenes per account completely free.
- JSON-Structured Prompts: Formatting image and video prompts inside JSON blocks prevents LLM hallucination and ensures rigid compliance with 9:16 vertical framing rules.
📌 Original Source & Attribution
- Source Video: How to Make VIRAL AI Pumpkin Fruit Videos FREE (Master Prompt)
- Original URL: https://www.youtube.com/watch?v=y84NhYM373o
- Credit: AI ZenStudios
