The information went relatively unnoticed, but last February, Coca-Cola signed a major partnership with OpenAI, the publisher of ChatGPT. The multinational firm intends to use AI tools for marketing purposes. In fact, these technologies are likely to revolutionize the industry in the future. In the meantime, the results are still in their infancy. Here are 3 attempts at false advertisements generated by an artificial intelligence.
Pizza Pepperoni
On April 24, the Internet user who goes by the nickname “Pizza Later” tweeted a video created by an AI. She is supposed to sell the services of an imaginary restaurant called “Pepperoni Hug Spot”. As we can see, people eating pizza have very disturbing distorted faces.
To obtain this surprising result, the twittos explains that he used GPT-4 for the script, Midjourney for the still images, and Eleven Labs for the voiceover. The rest was edited by him using Adobe After Effects software.
A fake beer
At the end of April, the advertising clip baptized “Synthetic Summer” turned a lot on social networks, and in particular Instagram. The latter is signed by the London production company Privateisland.tv. Helen Bower and Chris Boyle are the initiators.
According ArsTechnica, they reportedly resorted to Runway’s Gen-2 template which allows for short videos with simple written prompts. Stable Diffusion seems to be the tool used for still images.
Anyway, the result is again very “fake”, even if the barbecue evening atmosphere is reminiscent of many other advertising formats. But, something is wrong with this ad.
This inconclusive rendering is explained by the fact that video generators still lack a variety of sources, unlike image and text generators. But it is enough to remember the beginnings of GPT to understand that they could very quickly improve.
Harry Potter mannequin Balenciaga
Last month, user demonflyingfox posted a video on YouTube called “Harry Potter by Balenciaga”. The latter has already been viewed 9 million times at the time of this writing.
We can see the great figures of the saga transformed into models for the French ready-to-wear company. Even more disturbing, these characters look just like the models we see in fashion shows.
To achieve this result, AI Midjourney was again used for still images. The ElevenLabs tool was used to create the voices. The images were then donated to D-ID, a platform that makes it possible to create avatar videos, in particular by synchronizing facial expressions.
