If you asked me what the one big trend in generative AI this year has been, I’d point at video. 2022 was the year of the image (Midjourney – released summer 2022) and 2023 was the chatbot (ChatGPT – released Nov 2022), but there have been convergent trends around video this year that are worth exploring here.
Back in February, OpenAI announced Sora, and at the time I recall it being a ‘holy shit’ moment (defined by me messaging various people with links). Hollywood producer Tyler Perry cancelled a massive $800m investment in a sound stage upon seeing the announcement. Having worked in a digital video production company, this appeared to be a huge development to take advantage of. However, we are still waiting for this product release – almost a year on.
 |
| Sora was announced in February, but has yet to be launched. |
Possibly the delay was caused by numerous other generative video AI tools coming onto the market over the spring and summer. Dream Machine was released by Chinese developers Luma in June, while Kling AI arrived a month later.
Both of these seemed rather novel to me and I already had a Runway subscription due to its remove background feature. Runway had basic generative video baked into its toolset for a while, but it felt rather like Midjourney on its first release – kind of awesome, but not good enough to be useful. This changed in June when it launched its Gen-3 model.
Like all generative AI tools, testing and working with a system is an iterative and often time consuming process. What I was particularly surprised by with Runway was its fast generation. Resolution output and file size was still quite small, but nonetheless it was impressive. I then worked out that if you ‘storyboard’ your shots using Midjourney and then image prompt Runway, the results are all the more impressive. I made quite a few video generations over the summer, but have yet to publish them.
With Gen-3 Runway moved from being kind of awesome, but not useful, to really quite useful. I follow a lot of generative video AI experts on LinkedIn, and they use multi tool workflows to get the best end results. Even with these, I could not deem Runway to be production ready, but the potential was huge.
Several developments since the summer have shown the TV and film industry are taking AI seriously. Generative AI consultant Colin Byrne has wriiten an excellent list on LinkedIn, but what I saw was the following:
- ITV posted a job for a Head of Generative AI, but then appeared to retract it due to industry pressure. The job description was muddled, but I still feel this was a mistake.
- Disney have recently appointed a Head of AI and created a new division on how best to use it.
- George Strompolos (ex CEO of Full Screen), Jamie Byrne and Dave Clark launched Promise, an AI first studio with backing from Andreessen Horowitz. This will almost certainly be the first of many such developments.
- Coca Cola ‘reinvented’ their classic Xmas holiday advert using AI. Initially it was met with praise, but then turned to concern. I haven’t done such a deep dive to determine whether this generative AI version will make TV advertising, but I doubt it. It was a worthwhile test.
 |
| A clipping from the Wall Street Journal on A/H investment into Promise. AI first studios are clearly going to be a thing – who is qualified to create such companies is another matter, but this is the kind of team make up that will be expected by investors. |
All of this signals a big shift in how generative AI is being views in established media circles and how 2025 will shape up. Big production houses will be bringing in AI experts, new studios will be formed, and the advertising world (particularly the more nimble digital video advertising) could well turn on its head.
And we haven’t even reached the Sora release yet. My hunch is that OpenAI were happy to stay out of this year’s early fray, learn from others, then cause a massive splash when they are ready. It will likely represent a seismic change in the film and TV production industry.
