Imagine a vivid dream filled with mythical characters, and vast landscapes, and you as the protagonist in futuristic armour. Now, imagine describing that dream and watching it transform into a high-definition video. Sounds impossible? Not anymore.
Enter Veo2, Google DeepMind’s latest text-to-video platform, which creates realistic 4K videos from simple text prompts. Its capabilities have stunned the internet with mesmerising demo videos, leaving many in awe — and some video creators worried about their jobs.
But Veo2 isn’t alone. OpenAI’s Sora is another challenger in the fray, bringing its approach to this exciting new age of AI-generated video creation. So, how do these platforms compare? Which one is better? And what does this mean for the future of video creation?
Veo 2 vs Sora
Sora, launched in early December this year, is available to ChatGPT Plus users globally. While the OpenAI video maker has gained a headstart with general users being able to use the platform, Google’s Veo2 is still in its beta testing.
Seemingly, Google’s Veo2 has an advantage over Sora thanks to multiple reasons:
4K video resolution: Veo2 offers users video resolution up to 4K which means better quality videos. In contrast, Sora offers a maximum resolution of 1080p which isn’t bad but still 4K is 4K.
Video duration: Veo2 renders videos up to 2 minutes in length. Compared to this, Sora creates shorter videos of up to 20 seconds.
– Cinematic control: Veo2 offers virtual camera control with options for adding cinematic movements like pan, tilts, etc. Its accuracy has stunned many users online. You can even play around with the lighting for a particular scene. This helps enhance storytelling. Sora focuses more on style presets and storyboarding. It’s similar to editing photos on your phone. You can choose between each adjustment manually or use the presets that phones offer.
This video demonstrates the cinematic difference between the two. This tracking shot along a busy city street captured on Veo2 illustrates better results in terms of the camera angles and lighting as compared to Sora.
Or like how in this post, you can see the slow zoom-in shot or how the camera rotates around a stack of TV. Despite the prompt telling the camera to rotate, the output in Sora has rendered a static camera shot. This stifles a creator’s vision in many ways.
– Realism: Some Veo2 renders online show its true ability to output photorealistic videos. This is also true in the case of physics-based motion accuracy which makes animations look more natural. This is an area where Sora struggles.
The following video by X user Ruben Hassid shows the differences between the two video engines. There are several inconsistencies in the results Sora has generated whereas Veo2 has been able to render more lifelike results.
Ideally, Veo2 is a comprehensive choice for video creators, but sadly it’s not available to general users. Google’s DeepMind has made the tool available only to a select few users, with no clarity on when the final version will be rolled out.
Sora, on the other hand, is available for commercial use with a ChatGPT Plus subscription that costs around Rs 1676 in India.’