When OpenAI introduced Sora, it felt groundbreaking. However, today there are many alternatives that can generate equally impressive videos.

Runway ML’s recently introduced Gen-3 Alpha follows closely on the heels of Luma AI’s Dream Machine debut. Meanwhile, Kuaishou, a Chinese TikTok rival, launched Kling, a model that competes with Sora in its impressive capabilities, capable of generating videos in 1080p high-definition resolution at 30 frames per second.

Trained on both videos and images, Gen-3 Alpha will power Runway’s text-to-video, image-to-video, and text-to-image tools. It will also enhance existing control modes like Motion Brush, Advanced Camera Controls, and Director Mode, and support upcoming tools for more precise control over structure, style, and motion.

Just as OpenAI engaged with Hollywood, Runway has partnered with top entertainment and media to develop customised versions of Gen-3 Alpha. This customisation of Gen-3 models allows for greater control over style and character consistency, addressing specific artistic and narrative requirements. 

Unlike OpenAI, Runway has announced that Gen-3 Alpha will soon be available to everyone.

“Gen-3 Alpha catching up to Sora in 4 months means that Omni might also be caught up by September. Anthropic might even release their agentic AI before GPT-5 is released. OpenAI should either speed up its strategy or accept falling to second place,” wrote a user on X.

Sora is Way Behind 

One of the major drawbacks of Sora is that it is currently unable to generate photorealistic humans with consistent expressions and characters. However, Gen-3 Alpha excels at generating expressive human characters with a wide range of actions, gestures, and emotions. 

For instance, in the short film ‘Balloon Head’ produced by Shy Heads using Sora, they replaced the human face with a balloon to overcome Sora’s limitations. Walter Woodman, the film’s creator, said  that they edited the film using Premiere and After Effects. 

Notably, Adobe has recently announced that it is exploring partnerships with leading AI providers such as OpenAI’s Sora, RunwayML, and Pika. 

Moreover, Gen-3 Alpha is designed from the ground up for creative applications, which enables it to understand and generate a wide range of styles and artistic instructions. 

Apart from Gen-3 Alpha, Luma AI claims that Dream Machine, unlike Sora, can understand the physics of the world and how things operate in the real world. Another interesting aspect of it is that it can extend existing  images into videos. 

“Wow. The new model from @LumaLabsAI extending images into videos is really something else. I understood intuitively that this would become possible very soon, but it’s still something else to see it and think through future iterations of,” said former OpenAI researcher Andrej Karpathy. 

One of the other key differentiators is the photorealistic quality of its videos. The AI algorithms employed by Luma meticulously analyse and enhance every detail, from texture to lighting, ensuring that the final output looks almost indistinguishable from real-world footage. There are some limitations to Dream Machine, such as morphing, inserting text in videos, and camera movements. 

On the other hand, the Chinese Kling model can generate videos up to two-minutes long at 1080p resolution and 30 frames per second. Known for its photorealistic output and accurate simulation of real-world physics, Kling excels in 3D face and body reconstruction, making the generated content more lifelike and expressive.

OpenAI has a GPU Advantage 

A prime contributor to Luma AI’s success is AWS, which provided the company with the much-needed GPUs. 

“Great to see how AWS H100 training infrastructure helped the Luma AI team reduce time to train foundation models and support the launch of Dream Machine,” said Swami Sivasubramanian, vice president for data and machine learning services, AWS.

However, shortly after becoming publicly available, the website couldn’t handle the overwhelming demand. This is where OpenAI has an edge. As the developer of ChatGPT, OpenAI has access to Microsoft Azure, providing access to the latest NVIDIA GPUs. At Microsoft Build, CEO Satya Nadella announced that they would be among the cloud providers with access to NVIDIA’s state-of-the-art Blackwell GPUs.

OpenAI recently partnered with Oracle as well to gain access to more compute.

Will OpenAI Ever Ship?

There is another interesting story associated with Luma AI. At Google I/O, Google introduced its video generation model, Veo. However, due to security concerns, Google hasn’t released the model yet. 

Dan Kondratyuk, now a research scientist at Luma AI who  previously worked at Google, said that he left Google because the company wasn’t shipping any products.

“I left Google to join Luma. I was part of the team that developed Veo early on but knew it would never be shipped to the masses for quite a long time, just like Sora. Not until a company like Luma forces their hand, that is (at least I hope, gimme access),” he posted on X.

Meanwhile, OpenAI has adopted a strategy of announcing products to steal the limelight from Google without actually delivering them. For instance, when Google introduced Gemini 1.5, OpenAI announced Sora on the same day. The day before Google I/O 2024, OpenAI announced GPT-4o. However, its voice feature remains unavailable.

On the other hand, Google, much like OpenAI, has entered the race of announcing products. 

Recently, the company introduced its video-to-audio (V2A) model, which can generate audio for any video. Interestingly, this model can be paired with Veo to create shots with a dramatic score, realistic sound effects, or dialogue that matches the characters and tone of a video.

OpenAI should release Sora before it’s too late. The competition is not going to diminish. 

Hollywood actor Ashton Kutcher recently praised OpenAI’s Sora, saying that creators will be able to render a whole movie using it. “I have a beta version of it, and it’s pretty amazing,” he said. 

In a recent interview with The Wall Street Journal, CTO Mira Murati said that OpenAI is most likely to make Sora publicly accessible later this year.