Sora 2

Really impressive engineering work. The videos have gotten good enough that they can grab your attention and trigger a strong uncanny valley feeling.

I think OpenAI is actually doing a great job at easing people into these new technologies. It's not such a huge leap in capabilities that it's shocking, and it helps people acclimate for what's coming. This version is still limited but you can tell that in another generation or two it's going to break through some major capabilities threshold.

To give a comparison: in the LLM model space, the big capabilities threshold event for me came with the release of Gemini 2.5 Pro. The models before that were good in various ways, but that was the first model that felt truly magical.

From a creative perspective, it would be ideal if you could first generate a fixed set of assets, locations, and objects, which are then combined and used to bring multiple scenes to life while providing stronger continuity guarantees.

I wrote this a year or so ago: https://www.technicalchops.com/articles/ai-goes-to-hollywood...

"The studios and creators who thrive in this new landscape will be those who can effectively harness AI’s capabilities while maintaining the human creativity and vision that ultimately drives the art of cinema."

It is in many ways thrilling to see this come to life, and I couldn't agree with you more.

I'll have you know that in this year's Atlanta 48 Hour Film project (something I've been doing since I was a teen), several teams used AI.

Rewind to just one year prior -- 2024.

AI video was brand-spanking new. We'd only just gotten over the "Will Smith" spaghetti video and the meme-y "Pepperoni Hug Spot" and "Harry Potter by Balenciaga" videos.

I was the only person to attempt to use AI in 2024's competition. It was a time when the tools and infrastructure for video barely existed.

On the debut night, I was resoundingly booed by the audience. It felt surreal. Working all weekend to have an audience of peers jeering at you in a dark theater. The judges gave me an award out of sympathy.

Back then, image-to-video models really were not a thing (Luma launched "Dream Machine v1" shortly after this). I was using Comfy, Blender, Mocap, a full Mocap suit (the itchy kind), and a lot of other hacks to build something with extremely crude tools.

We lost a day of filming and had to scramble to get something done in just 24 hours. No sleep, too much caffeine. Lots of sweat and toil.

The resulting film was a total mess, of course:

https://vimeo.com/955680517/05d9fb0c4f (It's seriously bad - I hate it. It might legitimately be the very first time AI was used in a 48 hour competition.)

That said, it felt very much like a real 48 Hour competition to me. Like a game jam. The crude ingredients, an idea, the clock. The hustle. The corners being cut. It was palpable.

I don't think you can say there isn't soul in this process. The process has so much soul.

Anyway, fast forward to this year. Three teams used AI, including my own. (I don't think I have a link to our film, sadly.)

We all got applause. The audience was full of industry folks, students, and hobbyists. They loved it. And they knew we used AI.

The industry is anxious but curious about the tech. But fundamentally, it's a new tool for the tool box. The real task is storytelling.

already happening: https://x.com/ken_wheeler/status/1954343731579994593

What is the target market for this? The videos are not good enough for YouTube. They are unrealistic, nauseating and dorky. Already now any YouTube video that contains a hint of "AI" attracts hundreds of scathing comments. People do not want this.

Let me guess, the ultimate market will be teenagers "creating" a Skibidi Toilet and cheap TikTok propaganda videos which promote Gazan ocean front properties.