Which company excels in creating promises with large video models? Gen-3's demo effect is unparalleled compared to Sora!

06/19 2024 459

Awaiting open beta testing.

Since OpenAI announced Sora, the video generation field has officially pressed the industry acceleration button, with many domestic and foreign companies actively investing in research and releasing vertical large models specifically for video generation, as well as encapsulating their technology into AIGC products that everyone can use.

With the dramatic increase in the number of new players, the competition in the video generation field has intensified, with the biggest impact naturally falling on established competitors such as Pika, SDV, Google, Meta, and Runway, which released the third-generation video generation model Gen-3 Alpha yesterday.

Gen-3 is impressive, but you can't use it yet

Various demo videos released by Runway late at night showcase cinematic-level detail, directly shocking all netizens. Compared to the previous flagship video model Gen-2, Gen-3 has significantly improved in terms of model production speed and fidelity, while providing fine-grained control over the structure, style, and movement of the generated videos.

Runway stated that Gen-3 Alpha features high-fidelity video, precise motion control, realistic character generation, multimodal input, professional creation tools, enhanced security, and high-quality training. The training process of this model brought together the collective wisdom and efforts of researchers, engineers, and artists. It is this interdisciplinary collaboration that enables the Gen-3 Alpha model to understand and express multiple styles and film concepts.

The official demo video is 10 seconds long, with fine details in facial expressions and emotional portrayal in character generation, and the elements and lighting in scene and landscape generation do not seem out of place. Please note that the following content has varying degrees of compression due to conversion to GIF. If you want to see the original video, you can visit the Runway official website.

A woman rides a vehicle through a street with alternating light and dark, with natural changes in the external light shining on her face, and no discontinuity or out-of-place scenes appear in the vehicles passing by outside the car.

Source: Runway

A man seems to be watching a film in a dimly lit place similar to a movie theater, with highly realistic details in the reproduction of slightly red eyes, eye movements, blinking, and slight twitching of the mouth.

Source: Runway

A dilapidated house, with the ground magically transformed into a plant door, and plants swaying in the sun as the camera moves forward to reveal more details.

Source: Runway

A flame floats in mid-air, wandering down the street. The details of the flame are noticeably more difficult to grasp than other elements, with some drifting edges. Coupled with the sliding movements of people in the blurred background, this video exposes a weakness of Gen-3.

Source: Runway

Next is my favorite video, a cinematic shot that seemingly takes people into a vast otherworld. If I'm not mistaken, this type of shot is often used in movies like Jurassic Park and King Kong. The background of the shot is too vast, so I don't expect it to show many details

Solemnly declare: the copyright of this article belongs to the original author. The reprinted article is only for the purpose of spreading more information. If the author's information is marked incorrectly, please contact us immediately to modify or delete it. Thank you.