Will our human crews be replaced by bots? Or will AI become a useful tool within video production?
In the first of a two-part blog, Smart Digital editor Sean Miller (SM) and producer Mark Lenner (ML) discuss the latest in AI innovations, the upsides and the downsides of this technological game changer…
Huge promises are being made around the potential of AI to give individuals and companies the possibility to generate video content themselves. But like many new technological innovations, alongside these benefits come big warnings including a loss of revenue for creative industries and legal grey areas over ownership and usage. This week, we take a look at the ways AI can make the video production process run more smoothly…
What are the ways AI can benefit video production?
ML: AI is already helping us to become more efficient in terms of cleaning up the sound or to focus on specific audio, for instance, a speaker voice. With much of our work focused on events, this is particularly useful as these videos frequently feature a lot of background noise.
Using AI helps, not only to reduce this, but it can also save a lot of time as you’re running the audio through the AI. Plus, we can also use it to tackle issues such as colour grading.
What do you think is the biggest strength or potential for AI in video?
SM: At the moment, we use AI to improve the quality of a video, rather than replacing content. We’ll be able to do even more as the technology develops.
AI’s strength is in creating something using small pieces of information. For example, you might have a Broll shot which you need to make three seconds longer. You’ll soon be able to give your clip to the AI and it can generate the additional footage, taking into account the general motion and how the camera is moving. Of course, there will be limitations – with a crowd shot at an event, AI will not be able to generate everyone’s individual movements.
Editing is about finding a way to make the puzzle pieces fit. With AI, you have the freedom to change some of the pieces to be exactly as you want. You can generate a little extra of what you need here and there. The key here is to use AI to create a better edit, using it as a tool to create something of more value.
You might also be able to fix shots or clips which were previously thought to be unusable – for example, if they were overexposed, or too shaky, you can stabilize the shot without ruining it or cropping it too much. If the shot is too wide or you’ve lost the sky, for instance, currently, you won’t be able to bring it back but with AI, soon enough, I am sure it will be able to generate elements missing out of the frame.
Speeding up the process for adding subtitles
SM: The process of adding subtitles to a video can be speeded up using AI It’s really quick now because you can plug it into an AI and it will generate the file, add it into your edit programme and overlay it all.
This will still needs checking for mistakes, but using AI you can generate subtitles in different languages at the same time.
You’ll still need a native speaker to check it, but it will push you a bit further along with the process rather than starting everything from scratch and having to send a video to each translator.
With one video and subtitle for every language, the translator can watch it and make changes directly within the file sharing platform. It speeds the whole process up, because we don’t have to go backwards and forwards.
Social clips made simple
ML: At the moment, AI in video production is at the stage where it’s somewhere halfway between a self driving car and you still being behind the wheel. It can save you time, but everything still needs to be checked by a human being.
One feature which is starting to trickle into software is AI generated clips. This will be a great functionality, going forward, because once AI actually figures out which clips have the most meaning and impact from within a one or two hour recording session, it can generate this clips which you’ll be able to use pretty much straight away. At the moment, this is hit and miss. You might get a key quote, but most of the time the results are random clips, but the potential is definitely there.
For instance, when we’re on site we can pass on AI generated clips from the day or morning for social feeds which will save the team time, rather than someone sitting and going through the footage and picking out clips. Plus, it will free them up to be out on the showfloor or shoot getting more footage.