Until just a few days ago, there was something called Sora implemented on the ChatGPT interface. Then, without me noticing, it disappeared. Soon after, a new version, Sora2, appeared. This is OpenAI’s generative AI for video creation—but it’s invitation-only. Without an invite code, you can’t try it.

Yesterday, while attending a seminar, I learned something surprising: you can actually access Sora2 through Genspark’s AI video feature!

Trying Sora2 via Genspark

Here’s how I tried it:

Go into AI Video (not Super Agent). Click the option where you normally see Gemini Veo3. To my surprise, the label “Sora2” appeared!

Of course, I had to test it.

Introducing Our “Family Members”

For context (especially for overseas readers):

Copan is a plush toy that my wife and I treat as our son. Since we don’t have children, he has become like family to us. Bary-san is another plush, based on the official mascot of Imabari, Ehime (Japan). We have the small-sized version.

So, naturally, I asked Sora2 to let these two “talk” with each other. My prompt was pretty rough—something like “Have a fun conversation.”

The Results

The generated video took quite a while, but here’s what I noticed:

The two characters indeed spoke, but the conversation made little sense. The voices were so fast that I couldn’t follow them. I tried Genspark’s transcription, but the text didn’t line up at all. Honestly, YouTube’s auto-captioning works better. Interestingly, the video didn’t have the “Sora2” watermark. I’m not sure if that’s intentional.

So, the experiment was messy, but still fascinating!

Final Thoughts

This wasn’t a serious project—more of a detour. But the pace of change in generative AI is astonishing. Just when I think I’ve grasped one tool, another appears. Right now, I’m also struggling with Cursor, so it feels like we’re always running to keep up.

Still, watching Copan and Bary-san “chat” thanks to AI was a small joy. Even if the words didn’t make sense, the fact that it’s possible at all is amazing.