Grok Imagine First Impressions

A new version of Grok Imagine was released recently, and due to the mass excitement and hype surrounding Sora 2s release, it was largely ignored by the AI community.

Today, I decided to sit down and finally try it out as I’m always interested in seeing what each model has to offer.

I was pleasantly surprised in my tests, but yes, the model is quite censored (no surprise there!) The animations are much better than I expected, and if I could describe it in one word, it would be dynamic.

While Veo 3 follows prompts more closely, Grok Imagine seems to sprinkle in some of its own creativity, which makes for a treat when you do get your results. Generations are quick, and there’s a handy regenerate option if you want to try something a bit different.

One thing I’ve noticed is that it’s a little picky on image input sizes and may squash your images if they are too large, so you might have to resize them yourself. At least in my testing, the model does not seem to support text2image, but I’ve only been using it for the last day so I certainly could have missed it somewhere.

However, you can generate nice images right on https://grok.com/ if you don’t have any sitting around or you’re just curious what the model can create all on its own.

If I had to name a downside I’ve noticed so far, it’s that the sound is usually quite bad in that it is low quality or simply does not match the scene.

For example, I was making a romantic video and the sound was some very grating sounding (what I think was supposed to be) violins. It seemed more like something out of a creepy B-movie horror film than what I wanted.

Still, it didn’t detract from my experience since sound is easy enough to replace. The dynamic animations are more than enough for me!

All-in-all, it’s totally worth a try because it’s completely FREE! Although I have no idea if there are limits. I have made around ten videos so far and haven’t run into any as of yet.

Have fun, and I hope you like it as much as I did!