Using AI to Make a Creative Brief for Key Art
PART TWO OF A MULTI-PART SERIES EXPLORING THE USE OF GENERATIVE AI IN KEY ART CREATION.
Hello, and welcome back to my multi-part exploration of how AI tools might be used to improve the Key Art production process. Disclaimer: The methodologies I’m describing in this article were performed as a test to see how fully I could replicate a Key Art production process using AI and do not represent my official opinion on how all this should be done. I’m just having fun here :)
Picking up where I left off last week, I was giving an introduction into the Key Art Process and describing the problem space. Check the link here if you want a refresher.
This week, I decided to begin my experimentation at the beginning of the process, and work to use AI to create as much of what would need to be done as possible. I have some interesting findings to share.
Let’s explain what I mean by the “beginning”. I’ll copy/paste my brief description of the first step of the Key Art process below.
Phase 1: Briefing and Concept Development
Key Art starts with a creative briefing where the marketing team, directors, producers, and key stakeholders outline the project’s core themes, target audience, and desired tone. They may also specify where the art will be used (e.g., digital, print, etc.) and provide references for style (eg: a lookbook). Once the initial briefing occurs, mood boards and concepts are storyboarded out for review.
Let’s break down what this means in terms of deliverables for me:
A text description of the creative brief and strategic research.
A presentation deck that goes through the brief and includes references and imagery.
Moodboards and concept images
A presentation deck that summarizes the initial concepts being presented.
Phase 1 is actually a pretty gigantic phase. I can’t even finish going through it in this one article - I’ll have to tackle only the creative briefing and deck, and not the initial storyboards and concepts. I’ll save those for next week.
I wanted to work with a movie that really exists in the public domain, so that I wouldn’t need to make up too much of the plot myself. I also wanted to be able to compare the real key art for the movie with whatever I come up with by the end of the process.
After some digging, I found a great public domain movie to work with. It’s called Fear and Desire, and it’s Stanley Kubrick’s first film, released in 1952. Its plot walks through a war between two un-specified countries - and the absurdity and insanity of human nature during a war.
For my test, I decided to pretend that we’re going to refresh this movie by changing the setting to a future time in which rival mining companies are fighting over territory and resources on Earth’s moon.
For legal reasons, director, actors, and studios or production companies mentioned in the test brief will be completely fabricated. Please imagine that these are some of the biggest actors of our times playing these roles, and that they are also stakeholders.
OKAY! *Claps hands, rolls up sleeves.*
Let’s…get…into it!
Deliverable 1: Text description for my pseudo creative brief.
Tools used: ChatGPT and Google Search
Time to create: ~60 min with research time and revisions.
Result: I’ll give this a B. The writing of it is not bad, but I have to deduct a bit because of the challenges I mention below.
Opportunities: ChatGPT has the lowest-barrier-to-entry of any of the AI tools I’ve used, and it feels the most useful to an endless number of scenarios. It’s designed well and it’s fast.
Challenges: You can’t trust that what it’s giving you back is accurate even though it sounds legit. So you have to independently verify things, which takes time. Additionally, there are the concerns with security and privacy. The details of high profile film projects are scrupulously protected, so ChatGPT itself probably couldn’t be directly implemented without endangering confidentiality. It would have to be something like ChatGPT, but only accessible internally.
Deliverable 2: A visual presentation for my creative brief
Tool I used: SlidesAI.io
Tools I didn’t use: Beautiful.ai, Kroma.ai, Designs.ai, Zapier/ChatGPT combo
Time to create: ~60 min including research and testing different tools
Result: D minus. I got back something pretty garbage-looking from my test. I won’t give it an F because it did create the whole deck from just my prompt and it IS fully editable and it technically follows a design theme.
Opportunities: All companies, not just networks and studios, tend to do a lot of communication through professionally designed decks. So if there’s a way to speed up the process and make them more quickly, that’s great.
Challenges: I had to reject most of the AI deck creation tools because they weren’t actually generating a whole presentation with a text prompt - they were just offering some cool layout tools on a per-slide basis - and the whole process was still mostly by hand, individual slide by individual slide. So to me…that really doesn’t count as generative, because it’s actually just still you designing it and making all the decisions. (So you might as well save your money and just do it yourself). The closest functionality match I could find to what I wanted - which was the ability to interpret the text directly into beautiful slides that fit the theme of the text - was SlidesAI.io, which is a plugin available inside Google Slides. Unfortunately with this one, there’s just a long way to go from a design standpoint. It’s definitely interesting - but for now at least, I’ll be continuing to create presentations the ways I usually do.
I hope this experimentation was interesting to you. I can’t wait to keep going and show you more of my exploration process. As I mentioned before, I suspect that the best improvements to be made lie nearer to the end of the process rather than at the beginning, but I couldn’t resist the fun of giving the whole thing a go.
Until next time!