My Thoughts on Sora2 (Review)

OpenAI’s social AI video platform powered by Sora2 is all the rage lately in the AI community, and I was lucky enough to get a code to dive in and try it myself.

Immediately, I was impressed by how it is able to generate realistic (and suprisingly funny) shorts with very little direction.

Sora2 also knows many popular characters, and you’ll have no trouble finding a dozen or more Pokemon and Spongebob videos in the trending section in seconds. People have really made some creative and fun things, and it’s honestly really fun to scroll and see what others have done.

However, I wanted to see if I could push the tech further by using starter images (whether as img2vid or pseudo reference to video) — and it was at this point I started noticing some glaring weaknesses with Sora2.

That’s not to say Sora2 is bad by any means; the text to image capabilities are truly amazing and fun, but I feel that underneath all the hype, there is more that needs to be said.

I am no expert when it comes to video generation, I’m still pretty new and learning more every day with each new tool that releases. I’ve tried a plethora of video generation tools from Vidu, Wan, KLING, Veo and others — all of them have their strengths and weaknesses.

Sora2 is no different.

So where could Sora2 improve? Definitely in its img2video function.

Many times I will upload a starter image and my videos will fail to animate (no matter how I prompt) and backgrounds will remain static, clunky or otherwise really strange-looking.

Getting Sora2 to animate one of my images feels like winning a small lottery because it simply fails most attempts, and I have seen others running into the same issue. When it does work though, it’s awesome. 

I do hope OpenAI will continue to improve the model, and I’m certain that they will, but I felt that it needed to be said that Sora2 is not entirely without its flaws.

If you’re looking for something that does great with img2vid or has reference to video capabilities, I would say save yourself the frustration and try out something like Kling or Vidu instead.

But if you’re looking for an excellent meme-machine that understands comedic nuance? Sora2 is right up your alley.

It makes absolutely beautiful anime clips as well, and the lip syncing is impressive. I’ll definitely continue playing around with Sora2 as right now it’s free as long as you were lucky enough to get an invite code, so maybe I’ll learn some tips and tricks to make even better videos.

I think Sora2 is an awesome place to get started as a beginner as it’s very easy to navigate, and I’m always happy to see AI becoming more accessible to everyone, not just the most techy people. 

Rest assured if I see any amazing tutorials, I’ll be sure to share them on my X page, so follow me there if you haven’t already right here! → https://x.com/SORAY_AI 

I’m just a small hobbyist creator, so I’d love to make some new friends in the community. :)

If you’ve made any awesome creations, I’d love to see!

EDIT: So I saw a tip on Reddit earlier that I’ve been testing out and it seems to work…sometimes, but it’s still really inconsistent. Still, it has improved my results noticeably, so I want to share it.

Basically it’s just to make your starter image a reference sheet of your character(s) and go from there. In my experience, simple prompts are working better. A good amount of the time it still fails and just animates the reference image, but when it works it’s very nice!

Here’s what I’m using, just so you can get an idea of what might work for you! Happy generating, friends~