AI Shortform Content Generator Part 2: Concussion
Figure: Example frame from Google’s new Veo 3 AI video generator.
(A wide-angle cinematic drone shot gliding slowly over a fog-covered forest at sunrise. The camera pans forward through golden mist as soft light breaks through the trees. A narrow dirt path winds through the dense woodland, with dew-covered leaves glistening in the morning light. The air feels damp and quiet, only the distant sound of birds echoing. The style is moody and atmospheric, like a scene from a nature documentary intro. No people, no animals—just pure immersive nature, filmed in 4K with soft lens flares and shallow depth of field.)
In our week-long Instagram/TikTok trial, we found that AI video tools aren’t yet ready to fully automate content creation. Cutting-edge models like Google’s Veo 3 can produce near-cinematic visuals, but only in very short clips and only behind expensive subscriptions (Gemini Ultra is ~$249/month for full Veo 3 ).
Independent tests echo this: one editor spent $275 over a month on Veo 3 and concluded it was a “broken product” despite the hype. In practice, each video still needed carefully crafted prompts and human editing – the common refrain is that AI is a powerful tool but not a standalone solution.
Content on Instagram and TikTok went viral mostly when it was outrageous or funny (for example, a locally-shot meme video of a man on Clacton’s seafront being swarmed by ladybugs received strong engagement). However, this kind of sensational content often risks crossing lines. Platforms’ recommendation systems favor emotional, provocative posts, so the most-shared AI-generated clips were meme-like and edgy. That rewards quick virality but can clash with brand values. Many major brands shy away from controversy altogether, preferring to pay for guaranteed reach rather than gamble on unpredictable memes. (By contrast, small local businesses tend to build trust through community-oriented content, engaging in local events and stories often yields stronger loyalty than chasing virality)
AI-generated images and graphics were the bright spot. Modern image AIs (e.g. GPT‑4o’s diffusion models) now produce remarkable realism, making them ideal for things like YouTube thumbnails or illustrative social posts.
We plan to use AI to automatically generate creative images (thumbnails, overlays, stylized edits) which can then be refined by designers. However, we must also heed the risks: security experts report scammers are already exploiting this technology. For example, a viral demo showed how AI can turn a photo of a perfect car bumper into one with realistic “scratches and a small dent”. In fact, the FBI warns that fraudsters now weave AI-generated images and text into sophisticated scams. Deepfakes and counterfeit media are becoming so believable that subtle visual glitches, blurring around the mouth or jawline, odd lighting or motion artifacts, are often the only. In our own test posts, viewers even noticed “morphing” faces and unnatural camera shakes in AI clips, exactly as described by deepfake.
Key takeaways: AI is maturing quickly but humans remain in the loop. Next-gen video AIs can be impressive, but they come with premium costs and strict limits (short clip lengths, glitchy results). The social algorithms reward the riskiest, so any brand using AI content must balance engagement with ethics and safety. AI tools excel at generating base content (stock-like footage, draft scripts, meme formats) and at optimizing text (titles, captions, tags), but all output needs human oversight. Security-wise, organizations must stay vigilant: even though AI-driven disinformation (deepfakes, propaganda memes) has so far not toppled elections, it’s pervasive as. As one expert warns, AI misinformation “is here to stay” and companies should monitor it closely.
Our AI strategy at Pez Post-Production will reflect these lessons. We’re adopting a hybrid approach:
AI-assisted footage: Use generative tools to create raw video clips and animations (like stock footage) as starting material.
Meme-style content: Leverage AI to draft short, humorous clips or text memes – with clear guidelines to avoid sensitive topics.
Text optimisation: Auto-generate and refine social copy (post text, titles, descriptions, hashtags) using AI, then edit for brand voice.
AI-driven thumbnails: Produce and A/B-test thumbnail images from detailed prompts, comparing them against human-designed versions.
In all cases, humans review and polish every AI output. We are exploring further automation (scheduling posts, chatbots, etc.) but will rely on proven third‑party tools for now, always seeking the best solution for clients.
For the moment, the content already generated by this experiment will continue to run, and we’ll analyze its performance before making more changes. AI is evolving every day, so we will keep adapting our workflow. If you’re curious about our full social-media content services, check out our Social page at Pez Productions.