Accelerating your marketing ideation with generative AI – Part 2: Generate custom marketing images from historical references

This article demonstrates how to enhance marketing campaign image generation by leveraging historical campaign data with generative AI, using Amazon Bedrock, OpenSearch Serverless, and Lambda.

• System analyzes reference images using Amazon Nova Pro to generate descriptions and metadata
• Converts reference images to vector embeddings using Amazon Titan Multimodal Embeddings
• Stores embeddings in OpenSearch Serverless index for efficient similarity search
• Filters reference images by campaign objective and target audience for relevance
• Incorporates reference image descriptions into enhanced prompts for new campaign generation
• Uses Amazon Nova Canvas to generate final marketing visuals aligned with brand guidelines
• Bancolombia case study shows reduced iteration time and improved brand consistency

The solution enables marketing teams to scale content creation while maintaining consistency by systematically using past campaign successes to inform new visual asset generation.