Pixels & Algorithms: AI in Video Production 🤖📹
Guest post by Diarmaid Mac Mathúna
This month I’m delighted to have a guest post from Diarmaid Mac Mathúna - a fellow strategic communications specialist with over 15 years of experience in campaigns and marketing. He’s probably the person I know who has done the most deep thinking and developed the most interesting ideas around how AI is transforming the audiovisual landscape - so I’m really excited he has agreed to share some of his findings. If you want to know more, connect with him on LinkedIn or Twitter and check out how he works in EU communications here.
When people think of artificial intelligence and video, it’s often deep fakes that come to mind first. But AI is impacting the world of video in many different ways - only some of which are visible.
I’m going to explore the latest developments in AI video from three different angles that are relevant to communications and marketing in a European context: creating content, consuming content and copyrighting content.
So let’s dive in…
Transforming Video Content Creation
Automatic video transcriptions are being turbocharged by AI now, with specialist services like Rev being integrated into platforms such as Vimeo to enable the editing of video footage just by making changes to the text transcription.
Another really useful type of AI video tool are services like Opus Clip that will help you slice and dice longer widescreen content into short vertical videos for social media reels or shorts in minutes instead of hours.
The advent of multilingual auto dubbing, with added lip syncing, currently being prototyped by services such as Hey Gen will be a big game changer. It’ll mean that with a click of a button a video presenter or even an interviewee will talk to you in your own language. Spotify is also working on automatic multilingual versions of podcasts that sound just like the original narrator's own voice.
Services such as Synthesia and D-ID already enable this kind of language switching magic with their avatar presenters. But the so-called “uncanny valley” is still an off putting thing with these avatars and they just aren’t convincing enough to our human eye.
Where the whole world of AI video really starts to become mind boggling is when you start stacking generative AI tools to go from nothing to a video clip in a few steps either directly or from within platforms like Canva. You can use Midjourney or DALL-E 3 to go from just text to an image. Then upload that image to Runway to create a moving video clip. Then head over to an AI music app before finally getting ElevenLabs to read out your voiceover script. If you don’t have a voiceover script you can ask ChatGPT to write it. And if you don’t even have an idea for a video you can ask Daydrm AI to come up with one for you.
Below is a short video clip I created using AI tools - I used:
Midjourney for text-to-photorealistic image.
Runway for image-to-moving video.
Soundraw for customised ambient music.
HeyGen for text-to-speech voiceover.
Changing Consuming Video Content
The video content consumption side of things is also being transformed by AI in sometimes surprising ways too. We’re all well aware that AI algorithms are feeding our addiction to reels and shorts.
And when we’re watching videos in different languages we can start to benefit from the work around automatic translations that’s already happening.
But how we actually find videos on the web through search is changing too, and we can now ask Google’s Bard to help us find videos on YouTube. Google is experimenting with its search results that include embedded video clips - so we won’t even need to watch the full video on YouTube. Instead, Google will automatically show us the clips we need right on your search engine results page like a featured snippet.
With AI now able to effectively watch videos for us, we can even get long form online videos or lectures with automatic educational flash cards or quizzes to suit our own personal learning style.
Copyright Challenges
These rapid developments in the areas of video content creation and consumption are opening up challenging questions around copyright too. There are open questions about whether anyone’s copyright has been infringed in the training data fueling these AI models. It’s up to copyright holders to opt out of their data being used, but that’s a technical process and you need to know it’s actually happening to opt out.
The generated AI video content is also covered by different copyright laws that apply across EU member states, in the US and in the UK, and it’s really only case law that will clarify the situation about AI copyright. But what I’m hearing from multiple lawyers is that AI generated content cannot be copyrighted under EU law unless a lot of human effort went into its creation - and possibly not even then.
So we’re basically left with a situation in the EU where at the moment it looks like anything we create using an AI tool is copyright free. That sounds fine because it’d mean that you can use it freely. But it also means that anyone else can use what you create freely too - and you’ve no way to stop them re-editing your AI generated content for disinformation purposes because they haven’t breached any copyright.
In the run up to the European elections, it’s also important to note that the terms and conditions of some of the platforms such as Synthesia don’t allow the use of the generated videos in a political context. So to avoid the situation where you have to pull a video from social media, it’s best to read the small print first.
By factoring AI video into your communications strategies there are definitely big productivity gains to be made, especially around multilingual content and social media cut downs of longer videos. But deep fakes are definitely not the only reputational risks that communications people in organisations need to be aware of as they experiment with AI tools. Copyright and other challenges loom large and it’s important to get your AI video strategy right first.
Obsessed - great post!