Using AI in creative production
Using AI in creative production
When the winds of change blow, some build walls and some build windmills. This issue is about building windmills.
We can already use AI to produce concept art, 3D models, textures, audio, copy and creative ideas.
Looking at 2022 it feels like we've reached a turning point for exponential progress in AI mainstream adoption, I'm excited to see what surprises next year will bring.
As of today these are the tools and possibilities that have caught my attention.
1/ GPT
GPT is the tool which will have the largest impact in my day to day work. Here are a few examples of regular use cases :
producing copy and user research for the strategic elements of a pitch or client deliverable
generating idea listicles to proactively propose to clients
kickstart workshops and brainstorming sessions with more information, data or ideas
I've also begun using GPT to support technical workflows, our developers can learn and experiment with new technologies, or save time when optimising code by prompting GPT for advice (simultaneously improving performance).
And there will be wilder opportunities, such as using AI to power NPCs and character dialogues.
2/ 3D
A rapidly evolving space that has is becoming increasingly powerful, I'm curious to see what these tools will be able to produce in a year’s time. For now we've been able generate textures with Stable Diffusion or produce model variations to populate WebGL scenes.
We'll soon need to look further into text-based model generation with Nvidia or Dreamfusion.
Other possibilities that I have my eyes on for experimentation next year :
generating 3D models from images with Kaedim
create entire worlds in 3D with Promethean
3/ Creative
By far the most popular applications today, instantly generating incredible artwork.
Anyone can generate images with text using the Hugging Face demo, otherwise you can subscribe to Midjourney, Dall-e or Stable Diffusion.
I've also been having fun generating paintings from images with Ebsynth.
4/ Video
A tremendously exciting area of innovation for me although I have yet to test it on the next project. RunwayML drastically reduces production time on motion design and video, proposing a whole bunch of features to automate repetitive tasks.
Automatic subtitles, sound effect generation and beat detection alone would have saved me days on the last documentary.
5/ Audio
Sound is often even more intangible and we need to produce a lot of variations to agree on a direction. Looking at past experiences, we could save days of work using tools such as Soundful or comparables to :
generate and test narrative voiceovers
generate sound effects for WebGL scenes
generate character dialogues for games
Already in motion to experiment with these early next year.
6/ Pitching
In the waitlist to try this one out, but it looks like Magical Tome will make it into my pitch building process. Combined with MidJourney for visuals and GPT for research, copy and roadmaps I should be able to condense a day or two of work into a single hour.
Conclusions
The industry is going to shift dramatically in the next few years, I'll be doing my best to leverage these tools for creativity and output while reducing unnecessary work.
No doubt we'll need less hours to produce the same results, will agencies be expecting working hours to remain the same - coupled with an increase in projects? Or will we invest the extra time to stay relevant by learning new technologies.
Great read and some good suggestions. It's going to be quite interesting once these tools can be more personalized and persistent, and also once you can fine tune them to be subject matter experts in a particular area.