Short-form vertical video is the single most attention-grabbing format on the web right now, and a new startup is betting that generative AI can automate most of the work that goes into making it. The idea is simple: take a short script, produce a natural-sounding voiceover, and generate visuals that match the narration — all with machine learning — then stitch the pieces together into a publish-ready clip.
Why founders are drawn to short-form + AI
There are two things that make this an attractive project. First, short-form platforms give rapid reach. Reels, Shorts and TikTok-style videos are designed to be consumed quickly and can be discovered by audiences far larger than traditional social posts. For creators and brands, that means a single well-performing clip can drive a lot of engagement.
Second, generative AI has lowered the cost and technical barrier to producing content. Recent advances in text-to-speech, image generation and video editing mean a lot of the repetitive, time-consuming parts of production can be automated. Founders say combining those tools makes it feasible to scale output: dozens or hundreds of short clips a week rather than one polished piece a week.
Taken together, the format’s discoverability and AI-driven efficiency create a business case: high velocity content aimed at platforms that reward volume and novelty.
A typical AI-driven workflow
The workflow these startups outline is straightforward and modular, so different pieces can be swapped in and out.
- Script generation: either a human writes a short, punchy script (often 15–60 seconds) or an AI drafts variations based on a topic or a brand voice. Scripts are typically optimised for hooks and clear, short sentences. - Voiceover: the script is fed into a text-to-speech model that produces an audio track in a chosen voice. Teams often use expressive, natural-sounding models rather than robotic tones; some keep a human in the loop for intonation edits. - Visuals: this is the trickiest part. Options include AI-generated imagery or short clips, stock footage matched to the script, animated typography and simple 2D or 3D animations. Many approaches combine AI-generated assets with licensed footage and templates to get motion that feels dynamic. - Assembly and edit: an automated editor aligns visuals with the audio, adds captions and graphics, and formats the video for vertical ratios. Human review is common at this stage to catch awkward timing or nonsensical imagery.
The modular nature means providers can focus on one part — great voiceovers, say — or offer a full end-to-end pipeline.
Quality, authenticity and platform risks
There are obvious questions. Can AI make content that feels authentic and engaging? Not always. AI-generated visuals can look stylised or uncanny, and the small storytelling window of a 15–30 second clip leaves little room to hide weaknesses. Platforms and audiences reward authenticity, so overly synthetic content can underperform.
There are also ethical and policy issues: likenesses of real people, deepfake concerns, copyright and music licences, and disclosure of synthetic material. Startups expect to keep humans in the loop for vetting and to follow platform rules, but the standards are still evolving.
For now, the sweet spot seems to be fast, simple videos where novelty and volume matter — news snippets, quick explainers, listicles and promotional clips — rather than attempts to replace high-end creative work. If the tech and taste align, we’ll see more AI-first short-form content in feeds soon.
This article has been restored to the What's New On The Net archive as part of the site's relaunch.