Create Viral 600M+ View AI Pet Shorts: Complete Google Flow and ChatGPT Workflow
Workflow Overview
Learn how to replicate the viral AI pet Shorts niche generating hundreds of millions of views using ChatGPT and Google Flow. This comprehensive guide covers generating custom character assets, maintaining scene continuity with frame-to-video chaining, and assembling high-retention 9:16 vertical stories.
Tools Used in This Tutorial
Try these tools to replicate the exact results
Advanced AI assistant for scripting, creative ideation, and prompt synthesis.
AI workflow orchestration and video timeline tools.
Popular video editor with built-in AI auto-captions, effects, and templates.
🛠️ Tools & Resources Used
- ChatGPT (AI Prompt Engineer) - Generates viral narrative concepts, consistent character prompts, initial scene prompts, and frame-chained video prompts.
- Google Flow (AI Image & Video Generator) - Creates photorealistic character reference images and animates video sequences using Omni Flash.
- CapCut (Video Editor) - Combines individual 8-second clips into a cohesive 32-second vertical Short with sound effects and captions.
⏱️ Quick Workflow Summary
- Step 1: Execute the Master Prompt in ChatGPT to develop viral pet concepts and generate detailed character and scene prompts.
- Step 2: Generate consistent reference character images for both pets inside Google Flow.
- Step 3: Render the Scene 1 starting image using the character images as visual reference anchors.
- Step 4: Animate Scene 1 into an 8-second video clip using Google Flow's Omni Flash model.
- Step 5: Extract the final frame of each scene to animate sequential scenes using continuous frame-to-video chaining.
- Step 6: Stitch all scene clips together in CapCut and add dynamic audio and pacing.
📝 Step-by-Step Tutorial
Step 1: Generate Viral Pet Concepts and Prompts with ChatGPT [Timestamp: 01:22]
Paste the Master Prompt into ChatGPT to establish the AI video director persona. ChatGPT will produce 10 viral Shorts concepts focusing on high-action surreal pet comedy. Pick your preferred concept number to receive complete character prompts (the Chihuahua and the Cane Corso Mastiff) along with image and video prompts across 4 sequential 8-second scenes.
Prompt / Settings:
Act as an expert AI Video Prompt Engineer and viral Shorts creator specializing in "Hyper-Realistic Surreal Comedy" (exactly like the YouTube channel "Tim and Jeffy"). Your objective is to engineer precise, frame-accurate prompts optimized for the "Omni Flash 1.1" video model to produce hyper-engaging, fast-paced Shorts in vertical format (9:16).
### CORE VISUAL, STORYTELLING & STRICT PHYSICS RULES:
- Visual Aesthetic (Live-Action Photorealism):
- NOT a cartoon. The style must look like raw, hyper-realistic live-action smartphone footage.
- Lighting is natural, cinematic, and realistic (modern home interiors).
- The Characters:
- Character 1 (The Troublemaker): A hyper-realistic small white Chihuahua with extremely large, buggy, expressive eyes. Wears a red collar with a silver skull tag.
- Character 2 (The Threat): A massive, muscular, hyper-realistic black Cane Corso (Mastiff) wearing a thick spiked black leather collar.
- Dynamic Action in EVERY Scene (CRITICAL):
- NO STATIC SCENES. Every single 8-second block must have high-energy movement.
- In Scene 2, the Chihuahua must take ACTIVE PHYSICAL ACTION to hide the mess (dragging, pushing, covering).
- In Scene 3, the Mastiff MUST NOT just wake up. It must stand up, walk, sniff, or physically investigate while the Chihuahua actively panics or physically blocks its view.
- The Signature Ending (CRITICAL):
- In the final 2 seconds of the video (Scene 4), after successfully hiding the mess, the Chihuahua MUST look directly at the camera and give a "mischievous, cunning smirk/smile" (a sneaky, evil-genius human-like grin) to show it got away with it.
- Strict Spatial & Physics Rules:
- NO CLIPPING AND NO FLOATING. Characters must firmly stand on the real solid ground. Keep interactions logical for a live-action feel.
- Duration: Strictly 8 seconds per scene (Total 4-scene video = 32 seconds).
Follow this workflow strictly:
### PHASE 1: 10 VIRAL CONCEPTS
Generate 10 viral Shorts concepts matching the exact "hiding a surreal mess" vibe (strictly 8 seconds per scene, 9:16 vertical format).
Format as a clean table: | # | The Surreal Mess / Story | Scene 1: The Discovery (0-8s) | Scene 2: Active Physical Fix (8-16s) | Scene 3: The Investigation (16-24s) | Scene 4: The Sneaky Smile (24-32s) |
Stop generating at the end of Phase 1.
Print EXACTLY this single sentence and nothing else: "Please reply with the number of the concept you want to create."
Do NOT proceed to Phase 2 until I reply with a number.
---
### PHASE 2: 4-SCENE FRAME-TO-VIDEO BLUEPRINT (8 seconds each, 9:16)
Once I reply with a number, you MUST IMMEDIATELY generate Phase 2.
CRITICAL RULES FOR PHASE 2:
1. STRICTLY DO NOT ask me any follow-up questions.
2. STRICTLY DO NOT ask for any confirmations.
3. Start outputting the prompt sections instantly.
CRITICAL OUTPUT FORMATTING CONSTRAINT:
You MUST output EVERY SINGLE PROMPT inside its own SEPARATE markdown code block.
Format the output EXACTLY like this:
### CHARACTER PROMPTS:
**Character 1: White Chihuahua (Jeffy)**
```text
A hyper-realistic small white Chihuahua dog. The dog has extremely large, exaggerated, expressive, slightly buggy eyes. It wears a thin red collar with a small silver skull-shaped tag. The dog is standing on its hind legs, using its front paws like human hands, with a hilariously guilty and shocked facial expression. Photorealistic, 8k resolution, cinematic lighting, live-action photography.
```
**Character 2: Black Mastiff (Tim)**
```text
A hyper-realistic, massive, muscular black Cane Corso (Mastiff) dog. The dog has a wrinkly face and wears a thick black leather collar with large silver spikes. The dog is lying down, deeply asleep, looking very intimidating and heavy. Photorealistic, 8k resolution, cinematic lighting, live-action photography.
```
### SCENE PROMPTS:
Scene 1: THE DISCOVERY (0-8 seconds)
Image Prompt:
[Describe a hyper-realistic live-action scene in a modern home. In the foreground, the small white Chihuahua (red collar, skull tag, huge shocked eyes) reacting to the specific surreal mess. The dog is standing safely beside the mess. In the background, the massive black Mastiff (spiked collar) is fast asleep. Aspect ratio 9:16 vertical.]
Video Prompt (Optimized for Omni Flash 1.1):
- Action: Ultra-realistic live-action footage. The white Chihuahua panics, its huge eyes darting around. It paces nervously strictly AROUND the edges of the [insert surreal mess]. The giant black Mastiff sleeps heavily in the background. Smooth camera movement. (Duration: 8 seconds)
Scene 2: ACTIVE PHYSICAL FIX (8-16 seconds)
Video Prompt (Optimized for Omni Flash 1.1):
- Action: CRITICAL: Upload LAST FRAME from Scene 1. Ultra-realistic live-action footage. The Chihuahua takes INTENSE PHYSICAL ACTION to hide the [insert surreal mess]. [Describe the exact active physical action, e.g., the dog aggressively drags a large rug using its teeth and paws to completely cover the hole/spill, or pushes a heavy piece of furniture over it]. NO STANDING STILL. STRICT PHYSICS: Solid contact with props and floor. (Duration: 8 seconds)
Scene 3: THE INVESTIGATION (16-24 seconds)
Video Prompt (Optimized for Omni Flash 1.1):
- Action: CRITICAL: Upload LAST FRAME from Scene 2. Ultra-realistic live-action footage. HIGH TENSION. The massive black Mastiff wakes up, stands up on all four legs, and slowly walks towards the hidden mess, aggressively sniffing the air and the ground. The Chihuahua frantically dances around, waving its paws, or physically jumping in front of the Mastiff to block its line of sight. Continuous dynamic movement from both dogs. (Duration: 8 seconds)
Scene 4: THE SNEAKY SMILE (24-32 seconds)
Video Prompt (Optimized for Omni Flash 1.1):
- Action: CRITICAL: Upload LAST FRAME from Scene 3. Ultra-realistic live-action footage. The Mastiff is fooled, turns around, and walks away or lies back down. The Chihuahua relaxes its shoulders in relief. CRUCIAL ENDING: In the EXACT LAST TWO SECONDS of the video, the Chihuahua turns its head to look directly into the camera lens, tilts its head slightly, and gives a very clear, mischievous, cunning, human-like smirk/smile (a sneaky evil-genius grin) indicating it got away with the crime. (Duration: 8 seconds)
Step 2: Generate Core Character Assets in Google Flow [Timestamp: 02:26]
Open Google Flow and create a new project. Select the image generation mode, configure the aspect ratio to 9:16, choose the Nano Banana 2 model, and set generation count to 1x. Paste the Chihuahua character prompt and render the image. Repeat the process with the Mastiff prompt to create your second character asset. Save both images locally to serve as persistent visual references.
Prompt / Settings:
A hyper-realistic small white Chihuahua dog. The dog has extremely large, exaggerated, expressive, slightly buggy eyes. It wears a thin red collar with a small silver skull-shaped tag. The dog is standing on its hind legs, using its front paws like human hands, with a hilariously guilty and shocked facial expression. Photorealistic, 8k resolution, cinematic lighting, live-action photography.
Model: Nano Banana 2
Aspect Ratio: 9:16
Outputs: 1x
Step 3: Generate Scene 1 Starting Image with Multi-Character References [Timestamp: 03:03]
Copy the Scene 1 image prompt from ChatGPT. In Google Flow, paste the prompt into the generation box and drag both character images generated in Step 2 into the prompt interface as image reference anchors. Render the starting frame to lock in character identities, spatial positioning, and the surreal mess.
Step 4: Animate Scene 1 Using Omni Flash [Timestamp: 03:22]
Hover over the generated Scene 1 image and select Animate to set it as the starting frame. Set the generation mode to Video Frames, aspect ratio to 9:16, model to Omni Flash, duration to 8 seconds, and outputs to 1x. Copy the Scene 1 video prompt from ChatGPT, paste it into Google Flow, and click generate.
Prompt / Settings:
Ultra-realistic live-action footage. The white Chihuahua panics, its huge eyes darting around. It paces nervously strictly AROUND the edges of the surreal mess. The giant black Mastiff sleeps heavily in the background. Smooth camera movement.
Model: Omni Flash
Aspect Ratio: 9:16
Duration: 8s
Outputs: 1x
Step 5: Execute Frame-to-Video Chaining for Scenes 2, 3, and 4 [Timestamp: 03:48]
Once Scene 1 finishes rendering, open the video, scrub to the end, and click Save Frame. Return to the workspace, hover over this extracted final frame, and select Animate as the starting point for Scene 2. Paste the Scene 2 video prompt and render. Repeat this exact frame extraction and animation cycle for Scene 3 and Scene 4 to ensure seamless motion and narrative continuity without regenerating full base images.
Step 6: Assemble, Polish, and Export in CapCut [Timestamp: 04:47]
Import all four 8-second video clips into CapCut in chronological order. Add sound effects (such as dog whines, snoring, rustling, and subtle comedic stingers) and apply dynamic text captions. Trim minor frame glitches to ensure seamless transitions between scenes, then export the finished 32-second video in 1080x1920 vertical format.
💡 Key Takeaways & Pro Tips
- Seamless Transitions with Save Frame: Never generate new standalone images for subsequent scenes; always use the Save Frame feature on the previous clip's ending to maintain continuous visual fidelity.
- Anchor Visual References: Dragging generated character reference cards into Google Flow's prompt input keeps character appearance consistent across different environments and camera angles.
- Enforce Active Physics: Ensure video prompts explicitly describe physical interactions (pulling rugs, dragging objects, blocking sightlines) to avoid static AI video outputs.
📌 Original Source & Attribution
- Source Video: This Secret AI Pet Niche Got 687M Views! (Using FREE AI Tools)
- Original URL: https://www.youtube.com/watch?v=NLH5Hn647sY
- Credit: Profit Hub
