As a creative professional with 10+ years of experience in media workflows, I have gained insight into the many details involved in creating feature films, advertisements, YouTube videos, and other forms of digital media. This is my space for sharing creative workflow ideas and developments. Thank you for joining me in the Creative Workflow Lab.
When asked “What is generative AI?” OpenAI’s generative AI writing technology known as GPT-3 responds:
“Generative AI is a new type of AI that is capable of creating new data on its own.”
Simply put, generative AI technology can be used to create entirely original images, sounds, videos, or text. It creates realistic (and never-before-seen) data after learning from or “training” from a massive dataset. For example, Stability AI’s open-source generative AI known as “Stable Diffusion” was trained on over 2 billion images.
Remarkably, these AI systems are not directly re-using the training data, but instead are able to learn from the training data to generate their own unique data. There are lingering questions about the copyright status of media generated using an AI that was trained on copyrighted data. Competing AI platforms use different datasets and some may be in trouble if they acquired the training data without legal permission. Legal experts may be debating this issue for many years. Microsoft (who can hire plenty of legal experts) is investing in the generative AI space and is already integrating the technology into products.
It’s my understanding that generative AI functions like a human brain in the sense that it takes in information and then uses that as inspiration to create original content. Because of its enormous potential, I believe that no matter the cost, this technology is here to stay and will quickly become relevant for all creative professionals. At present, DALL·E 2 by OpenAI is being used to generate over 2 million new images every day. That number will surely increase in 2023 and beyond.
I’ve seen mixed reactions in the creator community, with some viewing generative AI as a threat to job security and others viewing it as the future of media production. As I said in my previous post on vertically integrated workflows, it’s my hope that the positives will outweigh the negatives. And there will be negative effects, some of which have already been widely reported. Either way, the cat is out of the bag. Generative AI is currently being developed and embraced by large corporations including Shutterstock and Adobe. There are versions of the technology that are open sourced and therefore harder to control. Some implementations could operate via blockchain technology and therefore will be difficult, if not impossible, to shut down. Despite its seemingly limitless capabilities, I don’t think generative AI will completely substitute the work of humans. For the majority of media workflows, I believe humans will be involved for many years to come. Humans provide AI with goals, guidance, and validation.
One thing is clear: creative work will increasingly be produced in collaboration with AI. This will dramatically change entire industries.
AI-generated writing tools will help with continuous script development. AI-generated images will become an essential part of every graphic design workflow. AI-generated audio will enable instant voiceovers and background music. AI video (that is, 100% AI-generated) is almost here. That will unlock even more creative efficiencies. I’m guessing generative AI will eventually be trained with volumetric data and have the ability to generate entire worlds. I think that realistic world generation is 10 years away. It could be sooner, we’ll see!
In the meantime, creative professionals will be expected to increase output without sacrificing quality. Generative AI will streamline the ideation, planning, writing, capturing, and post-production workflows. There’s been a Cambrian explosion of new AI tools, and it will take some time for them to be fully integrated. I’ve seen several mobile media production apps integrating generative AI. This makes mobile devices the ultimate all-in-one production tool. When the AI is doing the heavy lifting, you won’t need to be sitting at a desk to edit a feature-length film.
Creating media will feel like a conversation with AI. Humans write the initial concept as a “prompt” and then the AI generates a result. With the image generation tool DALL·E 2, you can erase a section of the first result and tell the AI to re-generate that specific area of the image. This allows for quick creative iteration. The back-and-forth reminds me of the workflow that a director of a traditional film production has with an editor. First, the director provides the vision and general notes for the project. Next, the editor presents the first pass. The director continues to give notes until it’s approved. In that example, the editor is AI and the director is you. And you’re not only limited to making a movie, you can make anything. That’s our future!
I’ll have more thoughts coming soon as this new technology evolves and is broadly implemented. Subscribe and share to support this newsletter, many thanks!
P.S. - Looking for a creative director? Let's work together: https://shanleo.com
Bonus AI-generated image for reaching the end! :)
