Creative Production

Embracing Gen AI: Getting started

Transforming The Landscape

“Synthetic voices are very much an alternative to real voices now, with intonation, speed, emphasis and all of the other things that are required in a decent VO completely achievable in AI”

At CRC we are starting to see some areas of real traction with Gen AI. There are references to live examples and some of the products available, but this is not an exhaustive list. We’re often asked by brands the question of where to start, so we conclude the article with some thought-starters.

Pre-Production

Brands such as Coca Cola, Land Rover, Vodafone and Original Source have all been making waves by using Gen AI in moving image advertising. For the more cautious brand or agency, there are ways to use AI in the pre-production process to great effect whilst other brands pioneer the way forward in the real-world markets.

Visualisation

AI is already being used to visualise creative and campaigns through both static imagery and pre-viz, speeding up processes, which in turn should be being passed onto brands in the form of cost savings. When it comes to creative inspiration, mock ups, shoot planning and storyboards, there is an interesting and powerful use case to think about.

The go-to AI tool in this space is Midjourney, however other platforms that are well worth checking out include Krea, Kling and Leonardo. For moving image, Runway is a good destination for visualisation, along with Kling, Minimax, Sora, Krea and Veo3 to name but a few.

Each is able to do some things better than others, and as they update they continuously leapfrog one another in their capabilities. 

Animatics

It is now possible to create animatics that are close to photo-real content which test more effectively. You can cast your characters, locations, lighting, and even produce accurate voice-overs and dialogue in a Gen AI animatic. For the brand adopting a wait and see attitude on legals, it is also risk free as it is only used as an internal research exercise.

Claire Randall will be speaking at the Marketing Procurement & Financial Management Summit in Frankfurt and Paris. Find out more HERE

It is significantly faster than the traditional animatic approach, at around 25-50% of the cost. The final bonus is that they test more effectively as the focus groups are seeing something much closer to what the final ad might look like.  For animatics the process is always to create a storyboard and use each frame as starting points for animation.

The tools typically used are Midjourney, Leonardo, Kling, Krea, Runway, Minimax, Sora and Veo (there are others on top of that). Prompters will move between them to get the best results. 

Testing

AI testing is now becoming a valid alternative or a complementary service to traditional focus group testing. Companies such as Daivid.co and Alison.Ai offer this service.

Broadly speaking, the insights are based on significant data points drawn from real panels of people who have been exposed to significant amounts of advertising to test their reactions. These then become the data sets that the AI draws on to predict reactions to the content that is being tested.

Research has shown it is statistically accurate in its speed, with almost instantaneous results. And at a fraction of the cost there is a clear use case for brands to use it to test quicker turnaround content or even more disposable social content that needs to be posted quickly in response to real world events. 

Production

For production there are already numerous applications for Gen AI.

Stills

It is now possible to create Gen AI imagery in any format – from small social posts up to 48 sheets.  This gives an alternative to stock imagery at similar price points; it allows brands to ‘extend’ campaigns efficiently where they may require more content or simply want an alternative to a traditional shoot; and it offers the ability to create stills for complex campaigns. 

Prompt artists rarely stay in one platform but instead move between them. Perhaps generating in Midjourney but improving on individual features in Leonardo before upscaling the image in Topaz or Magnific. Although most of the work is done in AI, there are elements of traditional post integrated into the process.

The market now has numerous LoRa models available, that allow prompt artists to create bespoke brand LLMs to re-create brand assets, which can then be manipulated into AI imagery. 

Moving Image  

There are now many new examples of ‘fusion’ adverts coming to market – a blend of in-camera and AI. The main area that will remain challenging for AI is around complex human performance which will take a lot longer to perfect. 

Virtual Production requires imagery for its LED screens, and this is a space in which AI has a lot to offer. Fusing AI and Unreal together will make Virtual production even more attractive to advertisers than it currently is as the cost of creating the backdrops (whether still or moving) will drop significantly and the degree of art direction that will be able to go into them will be higher.

However, the Holy Grail in this space will probably be the development of AI HDRIs. These are 360 degree images that are vital to AR environments. These will be a complete game changer for brands wanting to create immersive experiences for consumers, increasing the fidelity, yet decreasing the speed to market and cost of production. 

Music and Voice 

There are numerous AI Music services that give a viable alternative to library music. It is an interesting space to watch with services such as Udio.com and Suno.com worth a look. 

Synthetic voices are very much an alternative to real voices now, with intonation, speed, emphasis and all of the other things that are required in a decent VO completely achievable in AI.  It’s only a matter of time before actors and VO artists licence their voices for use in AI, but for the time being services such as Elevenlabs.io and Wondercraft.ai are worth checking out.

AI can now dub films, adjusting mouth movements to perfectly sync the language to the lips, bringing to life the combined potentials of these many technologies.

In-Housing 

One of the greatest areas of potential for AI is in the in-housing world. In-house operations are typically both producing original content as well as adapting and versioning. 

Over time it is inevitable that we will see an increasing level of automation in versioning and adaptation space where the AI is able to read, predict and create versions in any number of territories to any number of consumer types. 

Conclusion

So how can you begin to make sense of the Gen AI options available? 

• Talk to people or companies who are good at doing it and using it in a sophisticated way.
• Don’t get bamboozled by the tech. Whilst Nvidia and Stable Diffusion models are important to the platforms, what is important to you is how you can use the platforms to get what you need, not the tech that underpins them. 
• Beware the AI model (for creative work) that claims it can do everything without a human input. To get good results you still need humans who know their creative craft.

If you are ready to engage in a more hands on way we would recommend:

• Run a test e.g. animatics, stills, social content or internal comms work. 
  • Re-run some work you did traditionally in AI to compare the two
  • Work a live brief in parallel. Making one in a traditional way and the other using an AI approach
  • Or just run some real work in AI. 

About the author

Claire Randall is Founder & CEO of Claire Randall Consulting