Generative AI Won’t Go Back in the Box, so What Does It Mean for Post-Production?

I genuinely can't remember the last day I didn't at some point discuss or debate generative AI. It's everywhere. Is it coming for our souls or is it our saviour? Only time will tell but for what it's worth I thought I’d put pen to paper from the perspective of a creative post production studio.

Let me just set the scene. We’re an independent company headed up by a colourist, a VFX wizard and a producer/ex-director. So between us we know the challenges involved in production and post production having worked on anything from flying CG cars to multi award winning comedy series and everything in between.

Being boutique we don't have the rigidity of a larger facility and as such we always implement the latest technologies. This was true before AI and will no doubt be true after (if there is an after…). When generative AI was emerging, we experimented as we do with any other visual technology and continue to do so.

There are a couple of points that I think are always important to consider when looking at the future of generative AI in post production:

• It still takes time
• Technical corners can be cut but ideation can’t
• It's rarely perfect
• Generative AI is in need of a rebrand
• Is it stealing
• The cost
• Will it go back in the box

It still takes time. Producing work that has a generative AI element still takes time and knowledge. It is not a golden bullet that anyone can do. An understanding of what you want and a clear way of describing it is key. As an industry, and indeed society, will we all have to improve our writing or at least learn the language that best describes our vision for generative AI? Probably. Mainly because prompts will be everything in the future and already are now. Communication has always underpinned our whole industry and it is somewhat ironic that being able to effectively communicate with our AI assistants in the future will be essential.

These same technologies are affecting every other business in different ways but the net effect seems to be that everything just needs to happen faster. We live in a far more reactive creative industry than years gone by and as such everything needs to happen faster. Because of this, clients will still want someone else to handle work they don't specialise in. They simply can't take on other types of work too because their workload has increased. We have found this with clients actively asking us to make AI animations because they either don't know how or don't have the technical vernacular to describe what they want. Yes, it means that there are a whole slew of technical roles that will either be used less or redundant, but more will be created by needing the knowledge to prompt AI. Sometimes generative AI will absolutely be quicker, but there are also times when it won't be and understanding which path to take or combination of, is fast becoming a skill we’re specialising in. I like to think of post production a bit like a recipe as I'm an avid cook. If you're baking a cake you use a recipe to get the proportions right. In post production we traditionally had a fairly set recipe, now we have a myriad of ingredients and no fixed recipe so it's the skill of knowing how much of each ingredient to use to make the cake without any waste.

Technical corners can be cut but ideation can't. This is a bit of an obvious one but interpretation of our written descriptions is everything. Yes the generative AI models will improve but we can't write 'make something cool' and expect it to create a Cannes Lion commercial. In its current format there are loads that generative AI can do technically, that are tools in our arsenal to help our clients. Just staying on top of the sheer number of applications is almost a job in itself.

For any campaign that has a strategy (which there always is) then specific points in any production have to be hit. Whether that's brand recognition, logo placement, demographic. There's a lot of notes that need to be hit and creatives have to distil this into a narrative that we then help tell. The idea is still the linchpin. This of course could change and some brave brands may decide that speed is more important than thought and just prompt an AI tool to make a series of random scenes and whack a logo at the end. It is possible but I'm sure any good marketer will tell you that that's brand suicide.

Now where or who the ideas are being conceived by is a great question. Could in-house marketeers come up with the idea? Quite possibly and we’re seeing more and more brands hoovering up creatives in house so maybe that's a future, but the application of the ideas still needs to be done. Post houses have always been a hotbed of creativity and technology and so our ability to turn these ideas into reality will still remain, even if the processes change.

It's rarely perfect. There is an element of experimentation with AI. I’m yet to find anyone who has got exactly what they wanted the first time. When we have done a project that uses a lot of generative AI there is often a point where we start thinking that it would have been quicker to not use it all. It's easy to get pulled into a spiral of tweaking because it's not quite what you want. This is where traditional post production comes in and why we at Wash think there is space for generative AI in the production process. Creatives and directors may be trying for days to get a shot that does exactly what they want. Either because the prompt isn't specific enough or the interpretation is not what is expected, it doesn't turn out to be what was wanted.

We’ve found that using generative AI as part of the post process is the best use of it. Much like with normal shoots, a shot may be 80% there but the final 20% of clean up, removal, retracking, lighting etc can all be done in traditional post. The same is true of shots created using generative. Also a big issue currently (which no doubt will change) is the resolution. Most AI tools seem to top out at 720p HD. This means if you wanted to make a vertical crop for TikTok using the same footage, there would be serious pixelation. Again this problem can be resolved through combinations of traditional post production and generative AI working hand in hand.

Generative AI is in need of a rebrand. Generative AI is what we use in all the incredible tools today, it is not artificial general intelligence. For a more detailed understanding read this. Yes generative AI is a massive leap, but in many ways it's just a series of plugins. Yes some really clever plugins that draw from data or imagery larger than we can imagine, but plugins nonetheless. In an industry where brand recognition is everything, to me generative AI has unfortunately been tarred with a brush that always elicits a reaction (good and bad). Societally and economically times are tough and having new programs we don't understand, fundamentally changing our lives on what seems like a monthly basis, scares us. Doubly so when it's got a name that makes us think of Terminator 2, the Matrix or any other post-apocalyptic film in the last 100 years. I think personally we need to take the heat out of it. ChatGPT, Midjourney, Veo, Runway etc are all amazing but from our perspective they are plugins and stepping stones to help make productions faster and more efficient. They are not the end game.

Is it stealing? This is a tough one and will no doubt elicit debate. Now I'm not a programmer and I don't work in the AI space other than using it to help create productions so I can't speak with authority on the technical or legal aspects of it. From a moral standpoint I'm torn if I'm honest.

In one sense, knowing that an image I’ve just prompted (not created! It’s not creative) is an amalgamation of millions of images and data, many of which may not have been asked or have permission does make me very uneasy. And if we were just using that image and slapping a logo on it and selling it to our clients I would feel very uncomfortable indeed. However we never do that. Much in the same way as using stock imagery or stock CG models as a base and then working into them, adapting them and fundamentally changing the original generative AI element to the needs of a project makes me able to sleep at night.

For years prior to generative AI, we had clients asking us to heavily reference a look or style. Mood boards and keyframes have always been drawing influence from work that has been in the world previously. One can argue that generative AI in the creative industries is an extension of this process. Yes, the generative AI companies should absolutely be held accountable if they have broken the law to create their models. And yes, legislation needs to be stronger. But equally, for years we’ve been using the internet and happily posting photos and writing for all to see. Wanting our fellow humans to read our recipes or look at our holiday photos. Is this a question of whether we reap what we sow as a society? These are heavy and complex questions that I can’t answer but I do know that if I don’t use generative AI as part of a production process that all our competitors have a huge advantage over us so while it’s here we can’t not use it.

The cost. As a producer at heart, the cost of a process is a huge factor in it being used. The cost of using generative AI is a complex one. There's the direct cost to us, subscribing to a service (of which there are more everyday). There is no doubt that we have made savings over the years by speeding up productions utilising these services. On paper these services are undoubtedly cost effective. But as I've said before generative AI does not replace traditional post production. So if we need to create a background for a green screen shoot, yes we may use some generative AI elements, but we also need to use stock and put them all together using traditional compositing techniques. So more cost effective yes, but not cheap.

Then we have the actual cost. AI uses a lot of electricity. One generated image can use a huge amount of energy. More detail here. Using generative AI responsibly means reducing the number of iterations and changes we do. Now when you're in the depths of panic trying to generate an image for a pitch or a background element, that may be the last thing on your mind but bear in mind if you generated 50 images to get the one that works, that's enough electricity for your fridge for a day. Our responsible approach is to minimise iterations through effective detailed prompts.

Will it go back in the box? Do you remember those famous companies that took a stand when the digital revolution came out and refused to engage. They felt that their customers wanted to see them in person only, and experience their products face to face. They thought typewriters and the post man were more than enough for communication. No, no one does because they all went out of business.

Generative AI is most definitely now part of our lives. Debating whether or not it should be, is fun but sort of irrelevant at this point. It is here now, and has been for several years and getting better every day. As a forward thinking and progressive post production studio, we have to use all tools available to us to help our clients realise their vision. We’re genuinely excited by these new tools and as long as they are used responsibly and as part of a process to the betterment of a production we will continue to do so.

Ultimately they are another tool. Post production is made of creators, technicians, artists and producers. Post production at its heart is creative problem solving and how we solve the problems has just shifted. We have to shift with it.