A Closer Look at Haiper's AI Video Generation

We are at the dawn of the era of AI-generated videos. And with OpenAI announcing Sora recently, we are now seeing more video-generation AI surfaces. And one such AI is Haiper, which does things a little bit differently.

Haiper, a company started by two Deepmind alumni, Yishu Miao and Ziyu Wang, was initially focused on 3D reconstruction using neural networks. With time, its founders found video generation to be a more fascinating concept and began focusing on it six months ago.

Just like Sora and even any other AI model, Haiper AI takes in a prompt provided by the user describing a video. It then reads it, tries to understand the details in the prompt, and then generates something out of the information processed.

Unlike Sora, which can generate videos up to a minute long, Haiper can only generate videos two seconds long for HD quality and up to four seconds for videos with lower quality. And unlike Sora, anyone can get on their site and try it out for free. OpenAI hasn't released Sora to the public yet.

Haiper is taking a different approach with its model by making it free, as they believe it can help foster community building. And, also, they have the intention of open-sourcing their models, making them freely available for people to explore for different possibilities.

---

I had never tried to generate videos with AI before, so it's nice to see that there is now one that's available to explore and try some ideas.

The first thing it asks you to do is login. I chose the Discord login method. And then it takes you straight up to the dashboard with a few things you can do with the AI.

The first thing I tried to generate was this prop: a time machine activating and opening a portal to send a man through time in a car garage. And it generated this:

I didn't exactly have a picture in mind, but I thought it was very nice as it depicted the garage, the car, and a time portal well. It's something that I would expect in a movie.

Next, I asked it to generate a video using the prompt that Sora used to generate a video. I was curious to see how it would interpret the prompt and what it would make of it. Prompt: A movie trailer featuring the adventures of the 30-year-old spaceman wearing a red wool-knitted motorcycle helmet, blue sky, salt desert, cinematic style, shot on 35mm film, vivid colours.

Impressively, it generated something similar to what Sora generated with the same prompt. Here's the video Sora generated below in this tweet.

https://x.com/_olujay/status/1758525256199819270?s=20

Another thing you can do with Haiper AI is to animate your images. You simply upload an image to it and then have it animate it. A description can be provided to get better results.

I put in my photo of an American football and asked it to imitate it. Here's what it generated:.

https://twitter.com/i/status/1765036048902471693

I could go on and on with what I generated, but soon enough, after some trials, I had drawn some conclusions about it and its performance.

I mean, I have not tried Sora for myself yet, but I have seen what it was able to generate. In comparison with Sora, Haiper doesn't come close. Sora, for example, can generate up to a minute and deliver very realistic results. Haiper AI, on the other hand, can only do a few seconds and isn't even as impressive.

But then again, it is only the dawn of these video-generating AI models, and they will only get better with time.

---

There are a few other video AI models around already, but there's going to be way more around by the end of the year. This, at this point, is the lowest level we are going to have, as they will all get better and better.

If you would like to indulge yourself in a few tryouts, you can simply head on to app.haiper.ai.

---

_{Read More:}

_{Sora: The Dawn of an Era in AI Generated Videos}

_{OpenAI Bringing Common Sense to Their Models}

_{How Useful is the Most Responsible AI Chatbot, Goody-2?}