Hackweek: World's First AI Crafted Holiday Advert
Marketers can finally start to leverage AI, to create cost-effective content at scale.
As a marketer, creating high-quality, engaging content is essential to driving results. But what if there was a way to automate the process and create assets at scale?
Over the past nine months, we have seen exponential growth and adoption of artificial intelligence. AI-powered tools are becoming increasingly advanced, making it possible to generate creative assets at a fraction of the cost and time. AI tools can generate creative assets like images, videos, voiceover, and copy faster and more efficiently than ever before. The problem is creating consistent, on-brand content that is of high enough quality to use in campaigns. Traditional asset production is time-consuming, expensive, and difficult to scale, so we thought this was an important problem to tackle.
Introducing our latest Hackweek, Bob the Reindeer, our holiday advert created by AI. And, keep reading, we’re currently looking for 3 brands to run real-world campaigns in Q1.
What is AI, and how can it be used to generate creative assets?
AI is a form of computer technology that emulates the abilities of the human mind; AI learns from a set of inputted data and incorporates this knowledge into a complex decision-making system.
With the help of deep learning algorithms, AI can generate images based on specified rule sets or content generated by other AI platforms. While there are only a handful of AI platforms that exist:
https://imagen.research.google
https://stablediffusionweb.com
https://chat.openai.com/auth/login
There are hundreds of companies building products on top of these AI APIs, including:
https://hotpot.ai/art-generator
https://www.canva.com/features/ai-image-generator/
How to leverage AI to generate creative brand assets
AI has the potential to be a powerful tool for marketers looking to maximize efficiency, collaboration, and streamline creative asset production. While creating crazy single images of horses on mars is fairly easy to create, it's actually pretty difficult to create consistent, high-quality videos using these platforms. There is still a fairly large gap to be bridged if we are going to use AI as a reliable tool and system, including image quality, colors, context, and consistency.
That's the question our team set out to solve in our Hackweek. We knew the secret was, in effect, to leverage different AI tools to create a production workflow that would allow us to make content that could be leveraged by a brand. Taking the output from one AI platform and feeding it into another to continue the content creation workflow. All we needed was a subject, and since we started this at the end of November, the obvious piece of content that we decided to create was a classic, heart-warming holiday advert.
Rehab continually invests in research and development on new platforms and technology and how they apply to consumers. As is standard with all our Hackweeks, (and the clue is in the name) we tasked our teams with producing the work in 5 days to create the commercial. Introducing Bob the Reindeer!
(read to the end for the final video)
The content production pipeline
In most cases, a library or platform is needed to guide the operation of AI technologies. It pays to understand the underlying models and how the products interact with a responsible AI framework and datasets so that you can direct the AI properly.
We're reserving the nitty gritty details for our clients, but to give you an idea of the systems and processes we are using, we combined 5 AI platforms along with a creative and production workflow as follows:
- Rehab prompted OpenAI’s GPT-3, much like a creative director would brief their team, and got it to generate various story titles and full script options. After some optimisation, the final script was produced via a single tailored prompt.
- The script was fed into text-to-image generators Midjourney and Dalle-2 to create storyboards for the piece. Midjourney produced results with the best visuals, with Dalle-2 used to edit and refine images produced by Midjourney.
- Replica Studios was used to create a more human tone for the advert’s narration, while Soundraw contributed the orchestral arrangement, which sits beneath the voiceover.
To see what this means in real-life, try out this very simple hack that we sometimes employ for image generation.
Text to Image AI Generation
These platforms are AI image models that are trained to interpret text input, understand what the input represents, and then use complex models to generate original, unique visual content. It’s not hard to imagine that since the AI’s image output solely relies on the text you enter, entering more descriptive text will make a better video clips, with perhaps with more vivid colors.
A simple way to see this in practice is to daisy-chain two AI platforms together to make better images in a text-to-image platform.
- Create a free account or log in at https://openai.com/dall-e-2
- First, we’ll create a set of control images by asking Dall-e to generate an image, think of whatever you want, and type it in. Cool right?
- Now we’re going to make a better image, copy the text that you inputted into Dall-e and then visit https://huggingface.co/spaces/Gustavosta/MagicPrompt-Stable-Diffusion. Paste the description into the “Initial Text” input box
- Wait for it to generate the prompts, choose one you like, and now copy and paste the new description back into Dall-e
Compare the results, did this make a better image? Send your results to https://twitter.com/rehab_agency #aicreativity
The second AI model is a text dataset model that is designed to create more descriptive prompts. AI is totally random but in the extensive testing that we did, we found it was a superior result.
What this means for advertisers
Our Hackweek proves that AI can be orchestrated to create production-ready video content quickly and at scale. While our teams were hacking the technology and process, our strategy teams got to work to identify specific use cases for the technology;
Performance + Social
Marketing teams are required to create large volumes of assets for every campaign they run, catered to each platform they utilize with variants of each asset, to achieve better cut-through with their key audience types ultimately.
AI tools will make this process much faster and enable brands to optimize existing assets rapidly, and generate new assets on the fly to achieve the highest possible campaign performance.
Traditional advertising
Whilst the tech may not be ready just yet to replace high production formats such as TV spots (how long will it take to see a Superbowl video content produced by AI?), certainly, the ideation and storyboarding process can be dramatically accelerated through smart deployments of AI tools throughout production.
In times of stretched marketing budgets, with the spotlight on ROI more than ever, the workflow efficiencies these tools can provide make the business case for experimentation, in our opinion, a no-brainer. Other examples include explainer videos, training videos or other animated videos.
Where we're taking this
These tools are free or cheap to use; they don't require any specific coding experience or processes. We would encourage you to use them and familiarize yourself with the technology and how it works. The technology is only getting better, and this is not a moment to assume it's not right for you; complacency about these tools will not provide an unfair advantage to your company, client, or career.
We're an innovation agency that creates new tools and services for brands by leveraging new and emerging technology and understanding audience behavior. We're currently developing content production pipelines for consumer and B2B brands, including the development of a AI content creation platform.
To find out more about how AI content generation can accelerate your business and to take part in campaigns to launch in Q1, please get in touch with us; there are only 3 spots available.
https://www.rehabagency.ai/contact
Related articles
AI + Creativity: Leveraging AI the right way.
How Boost catapulted Turner Sports to pioneer NBA fandom through AI.
