It’s crazy to think about how far AI technology has advanced in the past year. Looking at generative AI video specifically, it was only one year ago that we all watched that one Wes Anderson Lord of the Rings video go viral.
Created by Curious Refuge, the video spawned a series of different Wes Anderson-inspired franchise trailers—and eventually a full-blown AI filmmaking course. Yet, when we look back at the original videos, they feel like they were from a lifetime ago at this point.
As we’ve seen with recent advanced AI video models like Luma’s Dream Machine, Runway’s Gen-3 Alpha, and the yet-to-be-released Sora from OpenAI, it once again feels like we’re on the verge of yet another major AI wave that’s going to push the technology even closer to what many have predicted will be replacement-level for creatives.
While we’re still not sure when (perhaps not if) that day will come, let’s take a moment to look at how far the technology has evolved in the past year by comparing the first Wes Anderson LOTR video to one just released from Curious Refuge.
A Look Back at AI Video One Year Ago
As you can see in the video above, which was uploaded on May 9, 2023 (which YouTube still calls “one year ago”), you can see why the trailer went so viral when first released. It’s not perfect, but it’s still very much a clear representation of what simple text prompts can create when strung together in a concise, directed manner.
In an interview with School of Motion, creator Caleb Ward shared that the original trailer was created using AI tools like ChatGPT, Midjourney, D-ID, and Eleven Labs for the full process.
When you watch the full interview on YouTube below, it’s interesting to see which tools have been quickly outdated as well as which methods are probably still the best to use today.
A Look at AI Video Today
Now, to give us some context as to how much AI video models have evolved over the past year, here’s the most recent version of the same video done with today’s AI video technology. It looks like Ward and the Curious Refuge team are using the same text and image prompts for the AI video, but letting the AI add in more advanced camera moves and motion.
The clip references that it’s using Luma’s Dream Machine, which is one of the most accessible of this latest wave of AI video models, and perhaps one of the most powerful as well.
However, as you can also see in both videos, while AI technology has advanced in terms of camera movement, motion, and facial expressions perhaps, it doesn’t look too different in other ways. What stands out most to me is that the characters look almost less real, or like animations themselves. This might have to do with the source prompts, but also could be indicative of just how hard it is to create life-like video.
What do you think of these two videos though? Let us know your thoughts on how AI is advancing in the comments below.
Author: Jourdan Aldredge
This article comes from No Film School and can be read on the original site.
