MiniMax AI Nails Human Movement With its New Text-to-Video Model
MiniMax AI is a latest ‘made-in-China’ artificial intelligence video generator having a text-to-video model.
While China is the world’s superpower in manufacturing, it is not much behind in everything else. In the AI scene, the latest ‘made-in-China’ product is MiniMax, an artificial intelligence video generator. Already, it has gathered a great rap for its realism, especially of humans.
The technical term for this type of AI is the text-to-video model. The world’s leading AI startup, OpenAI, also has a similar service named Sora. Among European startups, there is a trend to dismiss Chinese services as second-rate, but we think that would be a serious misstep.
Backed by multibillion-dollar corporations like Alibaba and Tencent, MiniMax is termed a unicorn startup, meaning its revenue crosses the billion-dollar mark. The companion AI app Talkie has seen over 15 million downloads. Much like Character.ai, users can create and converse with a virtual being.
MiniMax was announced through an official trailer shared on X. The video has been created completely through text-to-video prompts. The video is about a small boy who discovers a magic coin. Every time he touches it, he is transported to a different event in history. Realizing its magic, the boy decides to share his discovery with the rest of the world.
As we said, the selling point of MiniMax is its realism, expressly evident in the way it renders human gestures, an essential and particularly telling element of human interaction. However, this text-to-video service might not be as good as the marketing campaign paints it to be. But it is definitely on par with programs like Runway Gen-3, Dream Machine and Kling.
MiniMax Videos
The current model is the MiniMax video-01. It’s the latest offering from the startup of the models that generate speech, language, and music from a text prompt.
The founder of MiniMax, Yan Junjie said, “We have indeed made significant progress in video model generation, and based on internal evaluations and scores, our performance is better than that of Runway in generating videos.”
Given its amazing reception, it is hardly surprising that version-02 of the text-to-video model is already in the pipeline. Moreover, the plan is to bring in the feature to convert image to video, as well as a combination of text and image to video. The initial clip generation will also be made longer.
In videos, the quality is paramount. The generation resolution of MiniMax is 1280×70 at 25 frames per second. The camera movements can be described and altered in the text, just like Runway and Kling. The length of the video generation is only six seconds for now. As we mentioned, MiniMax intends to increase that length to 10 seconds to compete with the industry leaders in the next update.
Lots of people online have posted reviews from testing out MiniMax. We’ll try to sum up the gist here. The startup’s service is great. Perhaps we can compare it to Luma Labs Dream Machine. However, no matter what the CEO and their marketing team say, it’s not as good as Runway Gen-3.
Kling is another Chinese text-to-video service that is widely used in the West. This one is much better than MiniMax is various respects. For instance, it has a wider feature set which includes 10-second clips, and the Pro package generates longer videos and has image-to-video conversion too.
Well, MiniMax has promised that this is just the beginning and as the updates roll out, things will only get better. In the following weeks and months, MiniMax could turn the tide in its favor worldwide.
