- April 30, 2023
- Posted by: Jörg Forthmann
- Category: BEST PRACTICES
Stephan Zipperlen is Manager Audiovisual Communications & Formats at Covestro and has been an active member of the CommTech WG in the Technology cluster since 2022. Christof Schmid spoke with him about the current tools and trends around AI in video.
Christof Schmid: Everyone has been talking about AI since the beginning of the year, especially thanks to ChatGPT. What is the situation with videos from your point of view? Are there similar trend dynamics there?
Stephan Zipperlen: The pace and variety of possibilities also seem to be exploding in video this year – but AI has had a strong impact on workflows and developments in video for years – from “automatic” transcription, translations that retain timecode binding even in foreign languages, to auto-tagging from the metadata to videos. You could see this well for example in the quality of translation tools based on machine learning – through the models like GPT it sometimes feels that these aspects and possibilities are almost forgotten when talking about AI. Basically, videos are only about 25 frames per second, so the developments and possibilities we are seeing with images will sooner or later arrive here as well.
Christof Schmid: What added value do the new AI tools bring to video?
Stephan Zipperlen: They break down barriers – both in creating: in who can produce and edit videos, and in watching. For example, it’s getting faster and easier to transcribe and translate videos with tools like SimonSays. This is important wherever video needs to reach as many people as possible – in different countries, internal communications, some social media platforms, or office environments where video messages tend to be watched without sound. There are some AI-powered solutions and platforms that enable not only automation but also collaboration. At the same time, it is becoming easier and easier to produce videos – production is being democratized, so to speak. Many of the AI-powered tools could be considered No-Pro – so you don’t have to be a pro at shooting or editing video to get good results that meet the expectations of your corporate design or targeted channels. This trend is also greatly accelerated by apps like cowindow that don’t bring AI capabilities with them, and supported by AI with capabilities for storyboards or translations. This makes it increasingly easy to implement central specifications or coordinated storyboards with a smartphone and to produce videos in the corporate design.
Christof Schmid: Many tools seem to help make existing things better and faster. What’s new when it comes to AI for video?
Stephan Zipperlen: Absolutely – that’s an exciting question, and somewhat difficult – because ever since paint was used to “edit” on glass plates and negative video, every new tool has made video editing faster and better. What’s really impressive through AI is the jump through the media. Today, when I use editing tools like Descript in the transcript of a video to edit the text and the video message is cut to match directly, it’s a really handy tool. But the fact that I can now also edit texts, a slip of the tongue or an incorrect number, for example, and that the soundtrack is also corrected with the help of a learned, AI-generated voice of the person in the video, we are really entering new territory. This no longer requires an enormous amount of effort to learn like the first deep fakes, but becomes suitable for everyday use. From avatar creation and control to synthetic voices and virtual influencers, the variety of topics that AI is creating in video is growing and evolving at a rapid pace. Solutions like runway enable editing in videos that were previously felt to be reserved for images – for example, AI-supported slow motion can now be created from any video material, the image section in videos can be subsequently expanded, the graphics of the entire video can be transformed into different art styles, individual elements such as people can become marble or bronze statues, as known from Midjourney. For an AI, it seems child’s play to coherently post-process the 25 frames per second.
Christof Schmid: If I want a quick start in AI and video now – which tools should I definitely try?
Stephan Zipperlen: Definitely just that: Simply try it out. Because you don’t realize the many new possibilities until you dare to use all the new tools.
Without advertising individual platforms, I share a small list of examples of different solutions and platforms to try AI for video: Simonsays.ai, Descript.com, Synthesia.io, sketch.metademolab.com, runway.ml. For those who want to dive deeper, Futuretools or Futurepedia provide a great overview.
To see how editing video with AI support is transforming and opening up to non-professionals, a look at editing tools like Descript is exciting. And the next step when it comes to evolving video with AI is generative tools like runway: a recent example that opens doors to post-production and effects previously reserved for big budgets or experts, now available to all.
Christof Schmid: For many, AI currently means mainly ChatGPT. As a communicator:in, how can I use text AI to create better videos? Do you have a prompt tip or two?
Stephan Zipperlen: It’s fascinating to see what ChatGPT can do in such a short time – and even if you don’t forget that the content all has to be reviewed and adapted, some things can be worked out excellently. For example, I never tire of emphasizing the importance of video scripts that are optimized for “spoken word.” ChatGPT provides excellent tips and optimizations here.
For example, a good starting point would be:
>> How to optimize a video script for spoken language.
Another strength of ChatGPT is summarizing and condensing. So it can be good to create an “anchor phrase” for the beginning of the video that holds viewers’ attention – by creating curiosity, polarizing, or building tension. However, it is important to note here that all texts entered into ChatGPT potentially become publicly available. It is therefore essential to take data protection and data security into account.
One possibility for this requirement would include:
>> Create a polarizing/suspenseful first sentence as an introduction to a video with the following script to catch the attention/curiosity of the audience: “…”
Or also about:
>> Act as [famous video performance artist]. Describe a video idea for a short clip about [enter topic], follower by a script
Christof Schmid: The technology cluster of the Commtech WG is working on answers for PR practitioners’ own questions. What is the added value for you in your active participation in the topic cluster?
Stephan Zipperlen: The exchange alone is really enriching – just looking out of your own everyday work, out of your own company, and talking to others about the new tools and possibilities. And then, of course, the idea of opening up this world to as many as possible who have not had their own touch points with AI. In the cluster, we focus on exactly that: how to get started in the use of AI tools, recommendations, own experiences, but also to break down barriers or even fear of contact.