GenAI-Powered Influencer Ad Campaign Generator

• 

  Logo

• 

  Architecture

Inspiration

Modern ad campaigns are moving away from traditional approaches to highly personalized, data-driven influencer marketing. We wanted to automate this entire pipeline—from understanding a product, to discovering the right influencers, to generating tailored image and video ads—using cutting-edge GenAI and Google Cloud.

What it does

-> GenAI-Powered Influencer Ad Campaign Generator is an intelligent multi-agent system that:

-> Accepts a product description.

-> Extracts structured product details using Gemini.

-> Identifies relevant influencers via social media metrics.

-> Generates tailored ad copy, image, and video content using Gemini and Veo.

-> Stores generated assets in Google Cloud Storage.

-> Exposes a REST API for integration and demo.

How we built it

-> Framework: Google Cloud Agent Development Kit (ADK)

** Agents: **

ProductInfoAgent: Extracts structured info from raw description

SocialMetricsAgent: Mocks trending mentions

InfluencerDiscoveryAgent: Matches influencers based on product-influencer alignment

AdContentGeneratorAgent: Generates ad text, Gemini images, and Veo videos

Multimodal Models:

Gemini 1.5 Flash for text + image prompts

Veo 2.0 for video generation (Cloud-hosted)

Infra:

Hosted on Google Cloud Run

Assets saved to GCS

REST API via FastAPI + Uvicorn

Deployment: Dockerized and deployed to GCP, public endpoint for demo

Challenges we ran into

-> Session handling and async event flow in ADK needed careful orchestration.

-> Initial errors with ADK version mismatch, containerization bugs, and GCP health checks delayed deployments.

Accomplishments that we're proud of

-> Full end-to-end campaign generation powered entirely by GenAI.

-> High-quality videos and images tailored to influencers and product tone.

-> Successfully deployed and demo-ready via a simple API endpoint.

-> Clean modular agent architecture—scalable and production-ready.

What we learned

-> Deep dive into ADK's session, event, and runner lifecycle

-> Fine-tuning prompt engineering for better ad content generation

-> How to integrate multiple GenAI modalities (text, image, video) in a cohesive flow

-> Best practices for deploying fast inference pipelines on Google Cloud

What's next for GenAI-Powered Influencer Ad Campaign Generator

-> Real-time social metrics scraping from X (Twitter), YouTube, Reddit

-> Real influencer discovery and scoring engine using LangGraph + Open APIs

-> Frontend dashboard for campaign monitoring and visualization

-> Scheduled batch campaign generation for brands

-> Add watermarking and auto-upload to YouTube Shorts or Instagram Reels

Built With

• adk
• docker
• gcp
• gcs
• genai
• google
• imagen
• loom
• multiagent
• python
• twitter
• veo
• youtube