Higgsfield, OpenArt, Leonardo, FreePik. If you’re paying any of these platforms for access to AI video models you could reach directly, it’s worth asking what exactly your subscription is buying. Because ByteDance just put Seedance 2.0 inside CapCut for free, and the answer is getting harder to find.
Cast your mind back to 2024, when the AI video landscape was genuinely confusing. New models were appearing every few weeks. Each lived on its own platform, required its own account, operated its own credit system, and behaved differently enough that switching between them mid-project was a genuine pain. Into that chaos stepped a new category of platform: the aggregator. One login, multiple models, a single workspace. The pitch was simple, and for a while, it was convincing.
That moment has passed. The question now isn’t whether aggregators added value then. It’s whether they still do, and for most working AI video creators, the honest answer is increasingly less clear cut.
The case for going direct to the mothership has never been stronger, and three developments in particular have changed the calculation. Kling has become a serious, professional-grade platform in its own right. Runway has built an ecosystem that earns its subscription fee through genuine workflow thinking. And ByteDance, the company that built Seedance, has put that model directly inside CapCut, not as a separate tab, but as a native part of the editor where creators already cut, caption, template, and export. When the people who built the model decide to hand it to you themselves, what exactly is the aggregator still selling?
What aggregators actually are
Let’s be precise about what we mean, because the marketing is designed to make this sound more complicated than it is. An AI video aggregator is a platform whose core offering is access to models it didn’t build, wrapped in an interface it did. The value proposition rests on three things: convenience, some creative tooling layered on top, and the claim that the whole is greater than the sum of its parts.
Higgsfield is the most prominent example. The San Francisco-based company describes itself as an AI-powered video reasoning engine, a phrase that does a fair amount of heavy lifting. Its CEO has been direct about the underlying logic: rather than competing with the model builders, Higgsfield integrates third-party models into its platform. That’s the business, stated plainly. OpenArt operates on similar principles, offering a rotating selection of third-party models through a unified interface. FreePik, perhaps better known as a stock asset library, has extended into AI video through the same aggregation logic. Leonardo, which Canva acquired in 2024, is trying to follow a similar trajectory. The names change. The pitch doesn’t.
The tools that earn their subscription
The contrast with platforms that have built something genuinely distinctive is worth noting, because it sharpens the question of what an aggregator actually adds.
Kling 3.0, built by Kuaishou, isn’t an aggregator. It’s a single-model platform with its own generation engine, its own creative tools, and its own developing ecosystem. It’s also, bluntly, excellent. The generation quality is consistently among the best available, the interface has evolved rapidly from a bare-bones research tool into something that feels like a production environment, and the pricing is transparent. You know what you’re paying, you know what you’re getting, and there’s no intermediary taking a cut.
Runway has taken a different path, and it’s worth being honest about one that complicates the neat distinction between “tool maker” and “aggregator.” Runway now makes Veo 3 and Kling available alongside its own highly praised Gen-4.5 model (Seedance 2.0 is advertised as “coming soon”) which means it is, in part, doing what the aggregators do. The difference is in what sits around it. Runway’s platform includes over 30 production tools, from motion brush and generative audio to character consistency via reference images, upscaling, and a collaborative editing environment built for production teams. The third-party models are an addition to an ecosystem the company spent years building, not a substitute for having one. That’s a meaningful distinction, even if the lines are blurring.
LTX Studio, from Lightricks, occupies a third position. It’s building an open-source generation pipeline that gives technically capable creators direct control over their stack, without routing through someone else’s platform or credit system.
Each of these platforms has made a strategic choice: build the thing yourself, or build something genuinely distinctive on top of someone else’s thing. The aggregators, by and large, have done neither.
The CapCut moment
Here’s where the structural argument becomes concrete. AI Video Week confirmed that Seedance 2.0 is now available in the UK and other territories through CapCut and through Dreamina, ByteDance’s own AI generation platform. You don’t need an aggregator to get to it. ByteDance built the model, ByteDance runs the platforms, and ByteDance is handing it directly to creators inside a tool they already use for editing.
This isn’t an isolated development. It’s a signal about where the value chain is heading. When model providers distribute directly through platforms that already have massive user bases, the aggregator’s access advantage disappears. The convenience argument weakens. The only thing left is the creative tooling layer, and if that tooling isn’t substantially better than what the direct platforms offer, the subscription is just a markup.
The same pattern is emerging elsewhere. Google has Veo integrated into its own ecosystem. OpenAI, before Sora’s shutdown, was moving generation into ChatGPT. The direction of travel across the industry is toward direct distribution, not intermediation. It’s a bit like being a travel agent in 2002, still insisting that booking a flight is too complicated for normal people. It was true once. Then the airlines built websites.
When does the aggregator still make sense?
It’s worth being honest about who these platforms serve well, because dismissing them entirely would ignore what the market is actually doing. If you’re new to AI video and want to experiment across several models before committing to one, an aggregator removes a genuine barrier. If you’re a social media team producing high volumes of short-form content to tight deadlines, and what you need is a fast, reliable output rather than fine-grained creative control, the single-workspace model solves a real problem. Higgsfield’s growth tells you that millions of users find this compelling. That’s not nothing.
Where it gets more interesting is at the professional end. If you’re producing commercial work, if you care about the specific qualities of a particular model, if you want the creative tools that Runway or Kling have built around their own engines, the question becomes whether a single workspace is worth the trade-off in control. The further you move into serious production, the more that question matters. Aggregators clearly have a market. Higgsfield’s growth alone makes that obvious. The question for readers of this publication is a narrower one: is the subscription giving you something you can’t get by going to the source?
What to do about it
The practical advice is simple. Learn Kling. Understand what Runway is doing and why. Watch the CapCut integration develop. Know where LTX fits if you want to run your own stack.
And the next time a new aggregator lands in your social feed with a gleaming interface and a credit system that takes a moment or two to map back to actual pounds and pence, ask the question that the marketing never answers: what exactly is this built on, and can I get there myself?
Most of the time, the answer is yes.
