The Real Money in AI Video Isn't in Generating It

01 /The Concept

Rubbrband built exactly that: an AI-powered storyboard platform designed to replace those freelancers.

You feed the platform a script. Rubbrband's AI automatically breaks it into scenes and generates an image for each one.

The generated images can be edited directly with prompts – for example, "have this character hold a knife instead of a hammer."

Visual style is configurable too, from pencil sketches to photorealistic renders. Not that it matters much – modern AI video platforms can take any image as a starting point and apply a different visual style as needed.

A free tier is available with limited generations. The professional plan costs $49/month, though it's worth noting that unlimited shot generation per scene – the feature that makes it truly professional – is only available on the Studio plan, which requires custom pricing.

Rubbrband went through Y Combinator in the winter of 2023, and the concept has continued to evolve since – visible in its Product Hunt launch history. In July 2024 it was an AI image generator. By December 2024 it was a video platform with model switching and editing tools. Now it's positioned as a storyboard generator.

This is actually a healthy trajectory for a startup: start somewhere, then progressively find the problems most worth solving as understanding of the domain deepens. The result is a sensible MVP that's attached to a genuinely painful and in-demand task.

02 /Why It Matters

AI video platforms are becoming increasingly in-demand, because video is now the dominant format on the internet. These platforms broadly fall into two categories.

One category is AI platforms for "talking head" videos. HeyGen is the most prominent example.

You create an avatar – either designed from prompts or modeled on a real person – then feed it a script, and the platform generates a video with lip sync, facial expressions, and gestures. These videos can be fairly long, though the gestures and expressions cycle in noticeable patterns.

The other is AI platforms for narrative video – clips where characters don't just talk, but perform actions in a scene. Google's Veo, Kling AI, and Runway are the main players here.

To create this kind of video, you typically start with an image (generated or uploaded), then describe the character movements, camera angles, and other effects. The resulting clips currently run 5–10 seconds.

This length constraint is partly technical – AI still struggles to maintain visual consistency across longer sequences, causing character and scene appearance to drift. But it also reflects how professional video production actually works. Good cinematography is built on rapid visual change to sustain viewer attention. A scene is broken into shots, and shots cut quickly.

Consider a dialogue between two characters: one shot establishes the full scene, the next shows the first speaker in close-up, the next catches the second character's surprised expression, the next cuts to nervous hands, then to a shot over the shoulder, and so on.

A typical feature film contains 40–60 scenes, each broken into 10–20 shots averaging 4–6 seconds each. A film is essentially 1,000–1,500 of these short clips played in sequence. And it's exactly these short clips that AI platforms can now generate.

Breaking a screenplay into scenes, and scenes into shots, is the director's job in pre-production. On set, the director gets actors to execute the actions described in each shot while the camera captures it as specified.

In other words, a large part of what makes a film compelling comes not from the screenplay itself, but from the director's skill in translating that screenplay into a sequence of scenes and shots.

Current AI video platforms solve only the technical problem of generating those individual shots. A truly creative AI platform needs to handle two upstream levels:

- Turning a screenplay into a sequence of scenes and shots – ideally producing a starting image for each shot and a prompt for animating it, which a director can then refine to taste.

- Turning an idea into a screenplay – which a scriptwriter or the creator themselves can then edit further.

These two tasks are less technically demanding than video generation, but they're more creatively complex – because writing an engaging story can be approached in practically infinite ways, depending on the imagination and talent of whoever is doing it.

The AI video market will mirror the structure of the broader AI market. In the general LLM space, there are a countable number of foundation models from large players (Google, OpenAI, Anthropic, Alibaba, and others) requiring massive investment. Around those foundations sits an almost infinite ecosystem of tools and applications that use those models in specialized, differentiated ways. That application layer is where smaller companies make money.

The same structure will emerge in AI video: a handful of well-capitalized platforms that generate video from prompts, surrounded by an endless array of tools and apps that generate the prompts – in creative, specialized, context-specific ways. The real play is that clients pay for creativity first. And creativity lives at the application layer: the ability to produce a compelling story in the right voice for the right purpose, then feed it to a base platform to generate the final video.

That means a large share of the money in AI video will flow to "creative layer" applications that can produce engaging, cinematically structured stories – starting with tools that turn scripts into shot sequences, and eventually expanding to tools that turn rough ideas into full screenplays.

On that note: storyboarding – creating shot-by-shot visual breakdowns from scripts – is already a freelance profession. About 2,500 storyboard artists are active on Fiverr alone.

If freelancers are already doing it, it's ready to be automated.

03 /Opportunities

Video is the internet's dominant format, and AI video platforms will only become more capable and more widely used. This is a market worth entering now – to grow with it rather than scrambling to catch up later.

But the AI video market isn't just about AI video generation. And building a generation platform requires enormous capital: Runway has raised $544M for its narrative video platform, while HeyGen raised $69M for its technically simpler talking-head product.

The most interesting layer for startups is the application layer: tools that help generate the stories that then get turned into video using base platforms like Runway.

Even HeyGen is moving in this direction. The company has announced an upcoming "creative operating system" that lets users produce complete videos from a single prompt – which means internally it must be generating scene and shot sequences, otherwise the result is just more talking heads. Nearly 70,000 people are already on the waitlist.

That's not a threat – it's a signal. Universal platforms can't fully serve every niche. Specialized tools will consistently produce better results precisely because of their focus.

The path in: pick a specialization. Then build and refine a platform for generating compelling stories in that niche – and let third-party base platforms handle the actual video generation.

The immediate question for any founder entering this space: what kinds of stories are in demand right now, for which audiences, and what gives them structure? Nail that answer for a specific niche, and this space has real legs.