It was a Tuesday in the first week of February, my coffee was getting cold, and I was prototyping an empty state for a feature. The kind of moment where you want a small animation to give the screen some warmth. Normally I’d open After Effects, or I’d sketch something for a motion designer to build. Instead I opened Runway, typed a description, and had three options to look at in about four minutes. One of them was usable. I tweaked the prompt twice, regenerated, dropped the result into the prototype, and the empty state was done.

It was the speed that struck me. Not the quality, exactly, though the quality was honestly fine. It was that the cycle from “I need a small animated illustration” to “the animated illustration is in the prototype” had collapsed from days to minutes, and I’d done it without involving anyone else. That’s the kind of compression that, when it happens often enough across a team, changes what the team is.

By the end of February I was hearing variations of the same story from four or five other product designers. Not a survey. Just conversations. But the pattern was consistent enough that I started thinking about it as a real shift rather than a quirk of my week.

Runway’s drift into product design

Runway launched in 2018 as a creative platform with a heavy lean toward film and video production. The early use cases were rotoscoping, green screen replacement, and other tasks that would have required After Effects expertise and several hours of work. Useful, but a niche.

The model versions kept coming. Gen-1, Gen-2, Gen-3, Gen-4. By late 2025, Gen-4 was producing five-second clips with consistent characters, controllable camera moves, and physics that mostly held together. Filmmakers and ad agencies were the loudest user base. But the tool drifted, both because Runway shipped product features that made it easier for non-filmmakers to use and because the model quality crossed a threshold where short, specific outputs started being usable for purposes that weren’t film at all.

So what’s the actual shape of how product designers are using this stuff?

The use cases I’m seeing now fall into four buckets.

Onboarding videos. Short, looping animations explaining a feature. The kind of thing that used to either get cut from the project entirely or get outsourced to a motion designer with a two-week timeline. Now they’re a thirty-minute task at the end of the design phase.

Empty-state illustrations with motion. The case I described above. Empty states have been a long-running design problem because static illustrations age badly and motion makes them feel alive, but motion was historically too expensive to build for a state most users will only see once.

Reference for interaction design. This one’s subtle. Designers are using video generation to produce reference clips for what an interaction should feel like, then handing those references to engineers as part of the spec. “Here’s what I want the swipe to do, here’s what the transition should feel like.” The reference doesn’t ship. It’s a communication artifact, like a video Loom but more precise.

Marketing assets. Short product demo clips, hero animations on landing pages, social-media-format videos. The marketing team used to budget weeks for these. Now the in-house designer is making them in an afternoon.

The throughline is that motion has stopped being a specialized add-on and started being part of the regular design vocabulary, because the cost of producing it has dropped to zero or close to it.

The competitive field

Runway isn’t alone. The space is more crowded than it was even a year ago, and the tools have specialized in slightly different directions.

Pika leans toward shorter clips with more controllable composition. The interface is more designer-friendly than Runway’s, which started life as a filmmaker tool and still has that DNA. For product use cases that need quick, specific outputs, Pika is often faster.

Luma (specifically Dream Machine) has been pushing on physical realism and longer clip durations. For product design that’s less useful, because most product use cases want short and stylized rather than long and realistic. But Luma is the one I’d watch if you’re working on AR or 3D-adjacent product surfaces.

Sora is OpenAI’s entry. The quality is high, the access is uneven, and the workflow is more film-oriented than product-oriented. I’ve tried it for a few experiments. It’s overkill for product work most of the time and underpowered for actual film work. The middle position is awkward, but OpenAI has the resources to keep iterating until they find a more defensible product shape.

Kling and a handful of other tools out of China are shipping at a rate that’s hard to keep up with. I haven’t integrated any of them into my own workflow yet, but the demos suggest the gap with Runway is narrower than the Western press makes it sound.

For the specific job of “make me a short, looping, stylized motion piece for a product surface,” Runway and Pika are doing most of the work I see. For longer or more cinematic outputs, the picture is more contested. The space will look different in six months.

What this does to the motion designer role

I want to be honest about this because I have friends who do motion design as their primary craft, and the conversation about their role is a hard one.

The motion designer as a separate specialty inside a product team is harder to defend in 2026 than it was in 2024. The reason is straightforward: the work that used to take a motion designer a week can be approximated by a product designer with Runway and an afternoon. The approximation isn’t as polished. The motion isn’t hand-tuned. The timing is sometimes wrong in ways that a real motion designer would never let ship. But for a lot of product surfaces, the approximation is good enough, and “good enough on Tuesday” beats “perfect in three weeks” most of the time in product cycles.

That doesn’t mean the role goes away. It means it consolidates. Motion designers who can also do product design, or product designers who can also do motion design well, are more valuable. The pure-motion specialist on a small product team is in a tougher spot than they were a year ago.

The places where motion as a specialty stays defensible are the places where the bar is high and the cost of getting it wrong is real. Brand identity work. Title sequences. Featured marketing campaigns. Anything where the motion is the thing rather than an accent on something else. The motion designers I know who’ve shifted toward those higher-stakes contexts are doing fine. The ones who were doing in-product motion as their primary work are having a harder year.

The craft question

The honest version of this essay has to engage with the question of whether generated motion is actually good enough or whether it always needs hand-tuning to ship.

My answer, after a few weeks of using these tools as a real part of my workflow, is: it depends on the visibility of the surface. For a small empty-state animation that 5% of users will see for two seconds, generated motion is fine. The polish gap doesn’t really matter at that level of visibility and use. Ship it.

For a hero animation on a marketing site, the gap matters more. The animation is the thing the visitor sees first. Subtle timing issues, motion that feels mechanical, transitions that don’t quite land. All of those add up to a sense that something is slightly off, even if the visitor can’t articulate what. For those surfaces I still want a hand pass at minimum, and often want a real motion designer to build from scratch.

I’m not a motion designer, and I won’t pretend my reads on what works are as sharp as theirs. If you do this for a living and think I’m getting something wrong, please let me know.

The cheaper a motion piece is to produce, the lower the threshold for using it. The lower the threshold, the more motion ends up in the product. The more motion ends up in the product, the more the average product feels alive in ways static design never quite achieved.

For interactions inside the product (the micro-animations that respond to user input), generated tools aren’t really the right paradigm yet. Those need to be built in code with libraries like Framer Motion or native platform animation primitives, because they have to respond to runtime state. Runway can produce a reference for what the interaction should feel like, but the actual implementation lives in the codebase. That work is still product designer plus engineer, and I don’t see that changing soon.

What to try

If you haven’t integrated any generative video tool into your design workflow yet, the move I’d make is to try Runway or Pika on three concrete use cases over the next couple of weeks. An empty-state animation for a real product surface you’re working on. A short reference clip for an interaction you’re trying to communicate to engineering. A small marketing asset for a feature launch.

Pay attention to where the tool helps you and where you fight it. The places where you’re fighting are the places where you still want to involve a motion designer, or where the surface needs the bar that only hand work can clear. The places where the tool is enough are real time savings that you can route into other work.

You get the idea.

The integration doesn’t have to be all-or-nothing. The teams I’m seeing handle this best are the ones treating Runway as one more tool in the kit, not a replacement for craft. The tool is loud right now because it’s new. In six months it’ll just be part of how product gets made.

Jameson