The splash debut of Pika last week got me thinking about how AI tools are making real the promise of repackaging content for other media. Got an article? Run it through a multimodal AI (like Pika) and you have a video for TikTok or YouTube. Got a video? Go the other way and get the AI to write an article based on it.
In case you missed it, Pika 1.0 is a freshly-out-of-beta generative AI product from Pika Labs. According to the promo video (which you can check out below), the tool can create what look like cinematic-quality videos from simple text prompts. It also lets you perform what used to be complex edits — such as adjusting the canvas size and changing the clothing on people in the frame — in seconds.
Pika isn’t widely available yet (and the website has been crushed the past week with all the attention from the AI press) so I don’t know if it’s really that good at taking text and turning it into watchable videos. But there are any number of tools already available that do versions of at least some of what it promises: conjuring up video content for creatives who otherwise lack the skills, staff, or equipment to make it. They may not even have any visual assets in the first place.
That more or less describes any number of content operations, including this newsletter (though give me time). There’s a reason most online media companies tend to traffic primarily in the written word. The barrier to entry for text is the absolute lowest. Also, the internet is fundamentally a text-based medium. No one goes to Google and submits a video in the search box. By and large, you type text, and get back text.
That said, any SEO analyst will remind you that YouTube is the No. 2 search engine on the planet, and the lure of the online video ecosystem is huge for media companies if only for the lucrative rates they can charge advertisers. Yes, pre-roll and mid-roll ads are some of the most annoying on the internet, but that means they work — they capture the full attention of the viewer, and click-through rates are an order of magnitude greater than ads on articles.
Why ‘Pivot to Video’ Didn’t Work
That’s always been the fundamental science behind the “pivot to video” strategy that’s become a tragic media in-joke. You could write several books unpacking why that strategy was fundamentally flawed, but it mostly amounts to costs. Doing video content with any competence requires a minimum amount of investment in people (editors and producers), software, and access to raw video content. Even if you’re not shooting anything yourself, those costs are significant and they rapidly get unwieldy the more you scale. Add to all that the challenge of distribution (social networks and YouTube won’t cut it), and the ROI just doesn’t make sense for most publications. Many learned the hard way.
Now here’s generative AI, which appears to be bringing those costs down to near zero. You don’t need people to cut videos for you anymore; all you need to do is type. Want it vertical, short, with captions? Or how about widescreen, longer, and with an AI avatar? Just figure out the right prompts. You don’t even need to pay for an expensive Getty subscription for access to footage — the Pikas of the world will whip up synthetic visuals in seconds (presumably on licensed material, but we’ll see).
Putting aside what this means for jobs in video production, the rise of multimodal AI video generation will allow virtually any publication to, if not pivot to video, add the medium to its offerings.
Being able to offer video isn’t the same as offering quality video, of course, so a lot will depend on how good the tools turn out to be. And even if AI is competent at translating text into a video formats, any publication considering going this route would need to do some soul searching: Will the more abstract AI-generated stuff that a tool like Pika generates be appropriate for any publication covering real events? With the right style and templates, I can almost see it, but a lot depends on the audience any particular pub is trying to cultivate. That said, if there IS an audience for the content, someone will start serving it up. With costs so low, there’s no reason not to.
What seems clear, though, is generative AI is about to experience its own “pivot to video” moment. Whether that opens up real, revenue-driving opportunities for media companies is still to be determined, but at least this time it probably won’t be quite as pricey.
Like this post? Why not check out The Media Copilot podcast (listen on your favorite platform here, or via Apple Podcasts below)
Leave a Reply