CLONEI um Vídeo Viral de Restauração usando Agentes de IA (Perplexity + Nano Banana + VEO 3)

Introduction

This tutorial guides you through the process of using AI agents to create engaging car restoration videos, inspired by a viral ASMR video of a 1956 BMW. By following these steps, you will learn how to utilize tools like Perplexity Comet and VEO 3 to automate video creation without needing a physical workshop.

Step 1: Setting Up Perplexity Comet and Assistant

• Download and Install: Access the Perplexity Comet browser and install it on your device.
• Explore the Assistant: Familiarize yourself with the features of the Assistant, which will help you analyze and reverse-engineer video content.

Step 2: AI Analysis of the Restoration Video

• Select a Video: Choose a viral restoration video to analyze. In this case, we will use the BMW 1956 ASMR video.
• Input the Video URL: In the Perplexity Assistant, input the URL of the chosen video.
• Request Analysis: Ask the AI to "watch" the video. It will generate a detailed script and breakdown of scenes.

Step 3: Generating the Script

• Receive the Script: The AI will provide you with a script detailing 16 scenes.
• Review the Content: Go through the script to understand the sequence of events and visual elements.

Step 4: Creating Prompts for Image and Video

• Image Prompts: Based on the script, generate prompts for image creation. For example:
  • Prompt: "POV approach inside an old dusty barn, a heavily rusted 1956 BMW 503 grand tourer parked, flat tires, peeling paint, cobwebs and debris."
• Video Prompts: Create prompts for video generation as follows:

  {
    "scene_id": 1,
    "title": "Chegada ao carro abandonado",
    "duration_seconds": 8,
    "camera_style": "POV, caminhada lenta, leve handheld estável",
    "location": "celeiro antigo empoeirado, chão de concreto com detritos",
    "time_of_day": "tarde com luz suave entrando por frestas",
    "subject": "BMW 503 1956 extremamente enferrujada, pneus murchos, teias de aranha",
    "action": "câmera se aproxima lentamente do carro, revelando a condição geral",
    "visual_style": "4K ultra-detalhado, cinematic, contraste moderado, cores desbotadas",
    "sound_design": "som ambiente de galpão, passos suaves, ranger distante de metal, sem música",
    "notes": "ênfase na atmosfera de abandono; manter ritmo calmo e contemplativo"
  }
  

Step 5: Live Test of Scene Generation in Google Flow

• Open Google Flow: Set up your project in Google Flow.
• Input Video Prompts: Use the generated video prompts to create the first scene.
• Preview and Adjust: Run the scene generation and make necessary adjustments based on the preview.

Step 6: Plan for Future Videos

• Outline Upcoming Content: Based on the success of the first video, plan the next three videos in the series.
• Iterate on Feedback: Gather feedback on the generated content to refine your approach in future videos.

Conclusion

By leveraging AI tools like Perplexity Comet and VEO 3, you can create captivating car restoration videos without a physical workshop. This tutorial provided a clear path from selecting a video to generating and testing scenes. As you continue with this series, focus on refining your prompts and enhancing the visual storytelling of your content.