Runway Gen 4: Much Hype, Still Stumbling on the Basics

Apr 3 • News

Hey AI Filmmakers,

As a film editor who's had his hands on pretty much every AI video tool out there, I wanted to share my thoughts on Runway's new Gen 4 model that dropped last week.

The Good, The Bad, and The Reality

Let's be straightforward here. Runway Gen 4 brings some meaningful improvements to character and scene consistency across shots - something that's been a real pain point in AI video generation (VentureBeat). When characters shift appearance between frames or objects randomly change, it immediately breaks the illusion. Gen 4 does a better job handling this than their previous models.

But here's where I'm disappointed: for a fourth-generation product from a company that just raised $308 million in funding (TechCrunch), I expected more substantial improvements in the actual quality of video generation. The model still routinely misinterprets prompts, creating outputs that technically "work" but miss the creative intent.

How Does It Stack Up?

When I compare Gen 4 to competitors like Kling, Hailuo, and Google's Veo-2, Runway is falling behind in a critical area: accurately producing what you ask for. These other models often require fewer attempts to get something that actually matches your prompt.

Where Runway does shine is their control features. Their camera controls let you manipulate movement across six different axes (Stewart Gauld), giving you precise control over how your scene unfolds. But great controls paired with inconsistent generation quality is like having a professional steering wheel on a used car - the core performance still lets you down.

My Advice for You

For those of you working with these tools professionally:

Evaluate for your specific needs: Don't just go with Runway because it's the most talked about. Consider if its strengths align with your particular requirements.
Budget time for multiple attempts: When using Runway Gen 4, plan for several generation attempts to get results that actually match your creative vision.
Mix and match tools: Consider using Runway for specific technical shots where its controls excel, while using Kling or Hailuo for shots where accurate prompt interpretation matters more.
Be specific with prompts: I've found that extremely detailed prompts can help overcome some of Gen 4's interpretation issues, though it shouldn't be necessary.
Keep expectations realistic: Despite all the marketing hype, we're still in the early days of this technology. No model gets it right 100% of the time.

The Bottom Line

Look, I'm not saying Gen 4 is bad. It's just not the leap forward I was hoping for. Runway was an early pioneer in this space, and they've built an impressive ecosystem including Runway Studios for producing films and other media (Runwayml). But they need to focus more on improving their core video generation quality.

If you're already using Runway and familiar with its quirks, Gen 4 brings welcome improvements. The References feature for maintaining character consistency is genuinely useful. But if you're choosing between platforms, don't just buy into the hype - test how well each model actually follows your creative direction.

For now, I find myself reaching for Kling or Hailuo when I need reliable results, and only using Runway when I need those specific camera controls they do so well.

What's Your Experience?

Have you tried Gen 4 yet? How does it compare to other models you're using? I'm curious if your results match what I'm seeing. I've attached one of the more objective YT reviews below. I think Tim's opinion is pretty good, but you always have to consider none of these people want to completely alienate potential sponsors. Drop your thoughts in the comments.

Until next time,

Larry J.

Aiography.ai

3 comments