After first announcing a suite of AI capabilities last October, Meta is rolling out an expanded set of generative AI tools for advertisers. Advertisers now not only create different backgrounds for their product images, but also complete images that provide AI-inspired ideas throughout the photo, such as riffs that update the subject of the photo or the product being promoted. You can also request variations.
As an example, Meta modified an existing ad creative that featured a coffee cup next to coffee beans outdoors to show the cup from a different angle in front of the lush greenery and coffee beans. It shows how to conjure up memorable images. farm.
This might not be a big deal if the image was just to encourage someone to go to the local coffee shop. However, if what was being sold was the actual coffee cup, the AI variation provided by Meta could be a version of the product that does not exist in real life.
This feature can be abused by advertisers who want to trick consumers into buying products that don't actually exist.
Meta acknowledges that this is a potential use case, and advertisers can use the upcoming text prompt feature to display the generated output of their products in different colors, from different angles, and in different scenarios. It says it can be customized. Currently, the “different color” option allows you to trick your customers into thinking the product looks different than it actually is.
As the Meta example shows, the coffee cup itself can be made into different colors and viewed from different angles. Each cup features a unique swirl of frothy milk mixing with the hot beverage.
However, Meta claims that it has strong guardrails in place to ensure that its systems do not produce inappropriate advertising content or low-quality images. This includes “pre-guardrails” that filter out images that the generative AI model does not support, and “post-guardrails” that filter out generated text and image content that does not meet quality standards or is deemed inappropriate. included. Additionally, Meta said it stress-tested the feature with internal and external experts using llama images and a complete ad image generation model to uncover unanticipated uses and address any vulnerabilities found. Ta.
Mehta said the feature is already rolling out and will allow advertisers to provide text prompts to customize image variations in the coming months.
Image credit: Meta
Additionally, Meta allows advertisers to add text overlays to AI-generated images with a choice of 12 of the most popular font typefaces.
Another feature, Image Enhancements, will also be introduced in October 2023 and will be available in Reels on both Facebook and Instagram in addition to Feed. This option leverages AI to help advertisers adjust image assets for different aspect ratios, such as Reels and Feed. The idea is that advertisers can spend less time reusing creative assets on different surfaces. Meta said text overlays work in conjunction with image enhancements.
One of its advertisers, smartphone case maker Casetify, said it saw a 13% increase in revenue from ad spend by using Meta's GenAI background generation capabilities. The company is testing this option with its Advantage+ shopping campaign, and the AI feature first became available in the fall. The updated AI capabilities will also be available in Ads Manager via Advantage+ creatives, as before.
Image credit: Meta
Beyond images, Meta's AI can also be used to generate alternate versions of your headline in addition to your ad's primary text. This was already supported by leveraging the original copy. Meta says they are using previous campaigns as a reference and testing whether this text sounds like the brand's voice and tone. Text generation functionality will be migrated to Meta Llama 3, Met's next generation LLM (Large-Scale Language Model).
All generative AI capabilities will be available to advertisers globally by the end of the year.
Beyond the AI update, Meta also announced the expansion of its enterprise subscription service Meta Verified to new markets including Argentina, Mexico, Chile, Peru, France, and Italy. The service began testing in Australia, New Zealand and Canada last year.
Now, Meta Verified is offering four different tiers of its subscription plans, all of which include the basic features of verified badges, account support, and impersonation monitoring. Upper tiers include new tools like profile enhancements, tools for making connections, and other ways to access customer support.
Meta Verified will soon be extended to WhatsApp, the company said.