The Gist
- AI sameness is a human problem. Tools aren’t failing—creators are relying on shallow prompts instead of real creative process.
- Runway argues creativity still wins. CEO Cristóbal Valenzuela says distinct results come from vision, craft, and iteration, not button-pressing.
- Context is the real differentiator. Style bibles, references, and deep context inputs help AI generate truly original work.
Table of Contents
AI’s Sameness Problem
AI generated content tends to look, sound, and feel the same. OpenAI’s Sora videos blend into each other. ChatGPT writes with a characteristic style. Music generators like Suno generate what always feels like the same song, even across genres. This sameness problem for AI may be the single most significant issue holding the technology back from a wider explosion across culture today. But there may be a way around it. If you’re willing to do the work.
Why It’s Not the Technology
AI’s sameness problem is a product of the way people use the technology, not the technology itself, Cristóbal Valenzuela, CEO of video generation tool Runway, told me in an interview this week.
“If you’re not a filmmaker, I can give you an IMAX camera, the same camera that Christopher Nolan uses, and if I ask you to make a movie or shoot, you probably won’t match what he can do,” Valenzuela said. “And so that is not a representation of how good the camera is, the technology is. It’s a representation of the creative vision, or the artistic idea that you have behind it.”
Related Article: AI’s Sameness Problem: Marketers, Here’s Looking at You
The Creative Gap
Runway is a video generation tool valued above $3 billion, so Valenzuela’s fortunes are riding on its ability to churn out distinct content. It’s not as easy as writing a great prompt, he told me, and the technology’s capabilities are probably being underestimated due to the high volume of content from amateurs.
AI Opens the Door to Everyone
AI content generation tools have put the power to create in the hands of many people who’ve never thought through the creative process before. This itself is not a bad thing, and it may indeed lead to a proliferation of interesting new work from artists — professional and aspiring — who can create more with less effort and capital. But there’s a flip side.
The Hard Part of Being Original
“Being Christopher Nolan is very hard,” Valenzuela said. “It takes time. It takes iteration. Getting a camera and pressing record is very easy. Anyone can do it. I think AI is lowering the barriers of making stuff altogether. Anyone can make anything, right? So a lot of people make just bad stuff.”
To Valenzuela, the prompt is effectively the shutter: The final step after you think through the story, the art reference, the camera angles, and the nitty-gritty of filmmaking. “Pressing the button is the easiest part,” he said.
Before the prompt, Valenzuela said, diligent use of the context window is key to getting distinct AI content. This means uploading videos, photos, and references for the model to work from before it generates any content.
Most films also have a ‘style bible’ document that exists as reference point to help everyone involved in the production understand the art, direction, styles, story, and script. Valenzuela said AI filmmakers will upload their bible into the context window as well, giving it the details of what they’re aiming to produce. Then, the model follows their lead.
The Role of Iteration
With this starting point, AI filmmakers iterate over and over until the result feels right, and unique. “You iterate one thousand times, one million times,” Valenzuela said. “The best things I’ve seen made are people spending weeks just creating many variations of the same. And you pick the ones that are the best.”
Learn how you can join our contributor community.
