Where is creativity? [Horizon 01 - Issue #7]

March 10 - March 16, 2025

Can you do brainstorming with an AI? Can AI write creative copy? Can AI influencers be as “authentic” as human ones? Or is this just a peak AI efficiency game we are seeing?

We don’t know, but the signs are clear: AI is more capable than ever of creating truly creative content, previously a territory reserved only for humans.

As we approach the end of Q1 2025, I don’t think we have ANYONE left who questions this statement: “AI makes marketing more efficient with faster, better, and cheaper content production and marketing intelligence.”

But can we say the same for this one: “AI creates as good if not better creatives than humans, sparking true human emotions in those consuming this content.”

Because this is what creativity is. It sparks emotion. A great story, a sharply written copy, and an inspiring ad all share one thing: they make us feel something. Previously, only humans made other humans feel something.

Can AI replicate this? If yes, can it do better? Or at least, at a bare minimum, the same?

This week, we see examples of products willing to bet on the YES.

Us? We don’t know, but let us ask you this: holding a Mac, driving a sports car, or wearing a nice dress, do they make you feel something because you saw their marketing before or because they spark that feeling in their innate nature, as they are? It’s the chicken and the egg, right? We feel the same about AI stepping up as a creative partner in the AI marketing game.

Enjoy this week’s Horizon 01.

Peter & Torsten

AI-Produced UGC Videos (No Actors Needed)

Startup Captions.ai has unveiled Mirage, the world’s first foundation model for generating ultra-realistic talking videos in user-generated content (UGC) style.

With only a script, audio clip, or a short text prompt, Mirage can create a video of a person speaking – complete with natural facial expressions and body language – even though that “person” doesn’t exist.

This goes far beyond traditional lip-syncing tech. Previously, AI video tools could dub audio over an actor’s footage, often yielding awkward results (the lips move, but the face and gestures feel off).

Instead, Mirage generates the speaker from scratch so the voice, lip movement, facial expressions, and background align coherently. In short, it’s like conjuring a realistic virtual spokesperson to deliver your message.

Marketing in the Age of AI “Influencers”

The marketing implications are significant. Brands can now produce influencer-style testimonials or demo videos on demand without hiring talent or film crews. Captions highlight that Mirage lets you skip the usual months of outreach and contracting for live influencers – you can spin up a convincing promo video in minutes.

Need a peppy product review video for social media? Provide a script (or even let the AI generate one), and describe the on-screen persona (age, style, setting). Mirage will deliver a polished clip with a “person” enthusiastically talking about your product.

The model even supports 29+ languages, so marketers can localize content and reach different markets with authentic-looking videos in each locale.

Fast Creative Turnaround and Flexibility

The early use cases center on advertising. Mirage is rolling out via Captions’ Ad Studio, where brands can repurpose top-performing ads, create fresh concepts, and iterate quickly.

Imagine taking a winning ad script and instantly generating variations with different virtual actors or styles to see which resonates best—a new form of A/B testing for video creative. Moreover, the lack of real actors means unlimited edits and experimentation. Do you want the “spokesperson” to try a different tone or mention a new offer? Just tweak the input and regenerate.

This freedom, however, comes with a new consideration: ensuring audiences continue to feel a sense of authenticity. Marketers must balance the efficiency of AI-generated content with transparency and brand trust. Still, there’s no doubt that tools like Mirage signal a shift in social marketing, where the familiar face talking up a brand might just be AI-generated.

Snapchat Debuts Generative AI Video Lenses for Premium Users

Snapchat has introduced its first AI-powered video Lenses – filters that don’t just overlay graphics but actually generate new visual elements in your Snaps. Launched on March 12 for Snapchat’s $16/month “Platinum” subscribers, these new Lenses are powered by Snap’s in-house generative video model. In practice, they let users create short augmented reality videos with dynamic effects that were previously impossible.

For example, two initial Lens options, “Raccoon” and “Fox,” insert playful animated animals into your video in real-time, as if they’re right next to you. Another, “Spring Flowers,” magically zooms the camera out to reveal the user holding a bouquet of flowers that wasn’t there before. AI generates all of this on the fly, making the augmented reality experience more immersive and surprising than standard preset filters.

User Creativity and Brand Potential

Snapchat has a history of viral AR filters, and generative video Lenses take it further by allowing more open-ended creative outcomes. While currently a subscriber perk, it hints at where mainstream social content is headed.

Snap is rolling this out in a familiar format (the Lens carousel), lowering the barrier for users; people can apply the Lens and let the AI do the heavy lifting in creating the effect.

Implications for Marketers

Marketers should watch how audiences engage with these AI video Lenses. Early examples are lighthearted (cute animals, flowers), but they demonstrate that consumers are open to AI-generated enhancements in personal content.

Brands on Snapchat could leverage this trend by designing experiences that users will share – effectively turning fans into co-creators of marketing content. There’s also a storytelling angle: a travel company, for instance, could sponsor a Lens that “teleports” users to an AI-crafted vacation scene.

As generative AR tools mature, expect higher engagement on platforms that offer them since users get novel content to play with. For now, Snapchat’s move reinforces its position as an AR innovator. It shows how social media marketing might evolve with AI — more interactive, visual, and driven by user participation.

Google’s Gemini 2.0 Flash Speeds Up Creative Image Iteration

Google has updated its Gemini 2.0 Flash model to create illustrations, refine images through dialogue, and maintain consistency in characters and style across a series of images.

A standout feature is the model’s ability to handle conversational image editing, letting users tweak visuals through natural language in a multi-turn conversation. For marketers, this means rapid creative iteration: you can generate an ad mockup, then ask the AI to “try a different background” or “add our product name in the image,” getting updated versions in seconds.

Google reports that Gemini 2.0 Flash renders text in images more legibly than other AI models, making it particularly useful for creating advertisements and social media visuals with copy overlay.

This opens the door to quick A/B testing of ad creatives – for example, instantly producing several image variations (with different slogans, colors, or layouts) and gauging which performs best.

Accessible via Google AI Studio

Importantly, Google has made this experimental image generator free to test in AI Studio and through its API. Teams without deep technical expertise can try it via the Studio’s interface. By reducing reliance on manual graphic editing for each tweak, marketers can streamline campaign production and spend more time on strategy and storytelling.

OpenAI Previews a Creative Writing AI – A New Era for Copy?

OpenAI CEO Sam Altman revealed that the company has trained a new model geared towards creative writing, and it’s so good that it surprised even him. In a post on X (Twitter), Altman shared a metafictional short story generated by this unreleased model, saying it was the first time he had been “really struck” by something written by AI.

This is noteworthy praise, considering today’s top models (like GPT-4) are already quite advanced. Altman’s teaser suggests the model can produce fiction or narrative text with nuance or emotional resonance beyond current AI capabilities.

There’s no launch date yet, but OpenAI’s track record implies this model (perhaps a fine-tuned GPT variant or a new system entirely) could be the next significant addition to their lineup.

Impact on Marketing Copy and Brand Voice

For marketing and sales teams, a model that excels at creative writing could revolutionize content creation. Think campaign slogans, brand storytelling, social media posts, and even video scripts—all potentially written in a more human-like, engaging style by AI.

One immediate benefit would be speed and variety: a creative AI writer can brainstorm countless taglines or narrative angles in seconds, giving copywriters more ideas. It might also capture tone more effectively.

If this model truly understands stylistic and narrative “vibe,” marketers could use it to emulate a brand’s voice better or tailor messaging to different audiences (e.g., a playful tone for Gen Z vs. a professional tone for B2B).

Altman’s enthusiasm hints that the AI’s output felt less like a machine and more like a human storyteller, which could make AI-generated content more palatable for high-creativity tasks that humans have mostly owned until now.

Caveats – Creativity vs. Authenticity

While the promise is exciting, marketers should approach it with caution. Creative writing isn’t just about stringing nice sentences together—it’s about conveying genuine emotion and experience.

One analysis noted that today’s AI can mimic these qualities but doesn’t truly feel or understand them. In a branding context, an AI might produce grammatically perfect and even witty copy, yet subtle issues could emerge (a turn of phrase slightly off for the culture or a story that doesn’t quite align with the brand’s values).

The best use of such a model may be as a starting point or creative assistant. Human marketers could have the AI draft a few creative approaches and then refine the one that best fits the brand’s authentic voice.

If OpenAI’s new model lives up to the hype, expect a wave of adoption in marketing departments – but also expect savvy teams to develop guidelines on where human oversight remains crucial. This could usher in a new workflow where AI handles first drafts and ideation, and humans focus on editing, strategy, and injecting the true brand soul into the content.

An Opinionated Guide to AI Tools

This is not a news item but an article recommendation. It is easy to feel lost when deciding which model to use for what type of marketing work, which one is best for copy editing, and which one is best for brainstorming or image generation.

We loved Peter Yang’s “opinionated guide” to current AI models and agree with basically all his findings:

Claude for writing? It’s not even close, to be honest. ChatGPT’s Deep Research runs circles around Grok and Gemini. Perplexity is unbeatable in search accuracy (and speed), and Gemini is the best for image generation if you discount specialized models like FLUX.

Remember, no single AI model wins at everything, and marketers might benefit from using a combination.

Of course, this list might become outdated in a few days, and that’s the beauty of it. Our job with Horizon is to make sure you are always in the know. Until next week,

— Peter and Torsten