What AI Means For Video Production Today
The rise of AI and generative art tools has turned the video production industry on its head; now more than ever, there is a distinct uncertainty about what’s coming next. As the founder of a global video agency with 25 years in the industry, I run through what AI can and can’t do (yet) and offer actionable advice beyond the hype.
Setting the scene
Since its (debated) inception in 1888, the video and film industry has experienced no shortage of proclaimed ‘new dawns’; breakthroughs in technology which expand what we’re physically able to create, nurturing the cultural shift which grows alongside it. I distinctly remember when Jurassic Park first aired on screens in 1993; seeing the giant 3D-generated diplodocus lumber onto the screen was astonishing and cemented CGI as an invaluable tool in generating the impossible.
Seven years later and I’d go on to found Synima, a creative video agency that provides that very same 3D technology which enraptured audiences across the globe to commercial clients and brands. Seven years from the big screen to widespread commercial use is but a blink when you consider the wider scope of the industry (after all, it took sound nearly a decade to be embraced by filmmakers).
However, if the adoption of 3D animation from inception to commercial use is a blink, then the commercial journey of AI is a full-blown lightning strike.
From its humble beginnings as a crude AI-generated video of Will Smith eating spaghetti, AI has been developing at a rapid pace and is showing no signs of slowing down – taking on tasks that only months ago were deemed impossible. That same mystique and wonder people adopted towards the birth of 3D animation or the transition from film to digital is now accompanied by fear, confusion and uncertainty.
The Silicon Valley optimists among us might wave off concern as a pop culture hangover, chalking public scepticism up to the negative connotations artificial intelligence holds (read: Skynet, or the countless other media depictions of morally bankrupt robots set on destroying the human race).
However, with daily headlines announcing rather harrowing news around AI, such as the BFI stating “AI plundering scripts poses a direct threat to the UK cinema industry” (The Guardian), the paranoia is far more justified than the simple fear of the unknown that we’re led to believe. On the other end of the spectrum, there is huge excitement from brands and individuals alike looking to supercharge their content with AI.
For the first time, clients are telling us the tools they want used to make a campaign and the notion that AI is a one-click solution is still a prevailing sentiment.
Now more than ever, people are crying out for answers: what can AI actually do? Based on the innovative AI services work being carried out at Synima’s New York studio, I aim to dispel any myths and share the reality of AI’s actual capabilities across a variety of functions.
Text-to-image/video generation
Tools like Midjourney, DALL-E, and Stable Diffusion have made it remarkably easy to generate visual assets to match any desired style, look and tone. This rapid visualization of content is great for the initial concepting stage as multiple ideas (which would’ve taken days to create otherwise) can be explored instantly. This allows teams to try out ambitious concepts without much resource expenditure, often leading to more impactful creative without the risk.
In a similar way to its image counterpart, text-to-video generation allows the user a huge degree of creativity with little scope for consistency and alteration. Whilst great for quick concept visualization, most tools like Kling, Veo3 and Runway suffer from background motion limitations, inconsistent character appearances between frames and a brutal limit to duration (with clips typically capped at a few seconds).
Whilst easy to get stuck in, a lack of consistency between images and precise control over details renders AI generation potentially more arduous to manage for larger-scale projects with a need for rigid consistency. The sheer number of iterations required to get the exact shot you need not only leads to frustration and time delays but also adds an unknown element to your production: how many takes until I get the right generation? Will the tools I have available be able to execute the specific look I want? Will the servers of said tools be available when I need them? These are just some of the questions which get thrown into the mix when dealing with generative AI.
The wrinkles are fast being ironed out. By design AI systems are constantly learning and improving their output, whilst more complex systems like Comfy UI allow for greater levels of customization and character consistency. It’s only a matter of time before the aforementioned blockers become simple hurdles.
Voice & audio generation
Something as human and personal as a voice has also been the target of recent AI developments. Becoming remarkably more realistic as time goes on, AI voice synthesis tools can generate natural-sounding voice-overs or even clone specific voices (with permission).
Through multiple language options and the ability to adjust scripts and readings on the fly, this technology excels at making the voice-over process more streamlined and cost-effective. This isn’t without its limitations, however. We find many voice synthesis platforms lack the emotional nuance and correct pacing that are vital to a good voice delivery, often leading to an absence in authenticity and human connection – the last thing any creative wants when trying to connect to their audience.
With very little infrastructure for user direction and a proverbial minefield of ethical and legal issues, I find AI audio generation simply ‘not there yet’ when it comes to commercial viability.
Video editing & enhancement
Daniela Rus, Director of MIT’s Computer Science and Artificial Intelligence Laboratory stated in 2024 “Robots can be designed to filter the dull, repetitive tasks out of our lives” (The Heart and the Chip: Our Bright Future with Robots). It makes one consider an ideal reality – a world where the boring tasks of production become seamlessly automated and creative teams can work on the thing they care about most: creativity.
This is fast becoming a reality, with AI tools (many of which are integrating within traditional production software) now assisting with editing tasks like analyzing rushes, frame interpolation, upscaling and even removing objects from scenes. This takes one big step in realizing Rus’ vision: the use of AI to automate laborious technical tasks and speed up workflows.
Whilst I’m sure many editors are currently popping bottles of champagne at the prospect of never having to sync dailies again, there is a limit to how much we should leave up to AI in the post-production process. AI by design lacks the power to make context-aware creative choices, it doesn’t know when a shot has gone on for too long and it can’t distinguish between a good grade and a bad one. Relying too heavily on AI-powered decisions beyond mundane tasks can significantly hinder creative decision-making and make productions more sterile.
The hybrid AI approach
By now, you probably understand that AI is filled with “buts” – constant caveats and complications governing its use in the video production industry. AI is great for video generation but lacks scope. Voices can be replicated via AI but are devoid of emotional nuance. At every turn, there’s a trade-off.
This can quell some of the fears AI evokes but more importantly, challenges us to review our working relationship with it. We shouldn’t view this topic as a binary with pure AI as one option and traditional methods as the other, like any tool, AI’s usage should be based solely on a project’s requirements. When used correctly, AI opens a whole new world of unthinkable possibilities never before realized because of tight timelines and budgetary constraints. When used incorrectly, AI fosters the workflow equivalent of an Esher painting, an infinitely puzzling and utterly surreal exercise of the absurd. Much like the technological revolutions of old, the human element remains irreplaceable. A hybrid approach strategically combines AI software with traditional production expertise to deliver ground-breaking creative work while maintaining the human touch that drives emotional connection.
My Final Thoughts
The industry isn’t yet at the stage where we should stockpile tins of food and declare doomsday, but things are changing fast.
In the time it took you to read this article, there may have been a new update that addresses the limitations discussed – it’s that quick!
For now though, human influence and a genuine creative vision are still paramount to AI’s success, despite what social media and tech bro evangelists might have you believe.