ViGenAiR: Transform Long Video Ads into Shorter Formats with Generative AI

ViGenAiR (pronounced vision-air) is an open-source solution that uses state-of-the-art multimodal Generative AI on Google Cloud Platform (GCP) to automatically repurpose long-form Video Ads and generate several shorter variants, formats, and storylines. It generates horizontal, vertical, and square assets to power Demand Gen and YouTube video campaigns, leveraging Google Ads' built-in A/B testing to identify the best variants tailored to target audiences.

How it works:

• Frontend: An Angular Progressive Web App (PWA) hosted on Google Apps Script, accessible via a web app deployment.
• Backend: Hosted on Cloud Functions 2nd gen, triggered via Cloud Storage (GCS).
• Process: ViGenAiR uses Gemini on Vertex AI to analyze the content and storyline of a Video Ad, automatically splitting it into coherent audio/video segments. These segments are then used to generate different shorter variants and Ad formats.
• Components: Users upload or select videos via the UI, which are then processed by the Extractor service Cloud Function. The output includes transcription, video analysis, and coherent A/V segments, which are stored in GCS.
• User Controls: Users can steer the model towards generating desired videos via prompts and manual scene selection, and can generate, preview, and render video variants.

Benefits of using ViGenAiR:

1. Inventory: Generates horizontal, vertical, and square Video assets in different durations, allowing advertisers to tap into virtually all Google-owned sources of inventory.
2. Campaigns: Creates shorter, more compelling Video Ads that capture the meaning and storyline of the original ads, ideal for Social and Awareness/Consideration campaigns.
3. Creative Excellence: Produces coherent videos that follow Google's best practices for effective video ads, including creative direction rules for camera angles and scene cutting.
4. User Control: Users can steer the model towards generating their desired videos via prompts and manual scene selection.
5. Performance: Built-in A/B testing helps automatically identify the best variants tailored to the advertiser's target audiences.

To get started with ViGenAiR, follow these steps:

1. Fulfill Prerequisites:
   • Ensure your system has up-to-date installations of Node.js, npm, the gcloud CLI, and git.
2. Clone Repository:
   • Use git to clone the ViGenAiR repository.
3. Navigate to Directory:
   • Navigate to the directory where the source code is located.
4. Run npm start:
   • Execute the npm start command.
5. Enter GCP Project ID:
   • Provide your GCP Project ID and choose whether to deploy GCP components, the UI, or both.
6. Authenticate:
   • Authenticate to both Google Workspace (via clasp) and Google Cloud.
7. Deploy Components:
   • The script will create a GCS bucket, deploy the vigenair Cloud Function, and deploy the Angular UI web app to a new Apps Script project.
8. Access Web App:
   • The URL of the web app will be output at the end of the deployment process, which you can use to run the app and start generating videos.

To use ViGenAiR, you need the following:

1. Google Account: Required to access the ViGenAiR UI.
2. GCP Project:
   • Vertex AI API enabled: Required to access Gemini in Vertex AI.
   • Video AI API enabled (Cloud Video Intelligence API): Required for analyzing input videos.
   • Roles:
     • Vertex AI User role for all users running ViGenAiR.
     • Storage Object User role for all users running ViGenAiR.
3. Additional Roles for Deployment:
   • Storage Admin for the entire project OR Storage Legacy Bucket Writer on the <gcp_project_id>-vigenair bucket.
   • Cloud Functions Developer to deploy and manage Cloud Functions.
4. Components Created by Setup Script:
   • A GCS bucket named <gcp_project_id>-vigenair.
   • A Cloud Function named vigenair for Extractor and Combiner services.
   • An Apps Script deployment for the frontend web app.