What is Sora like and what are the rivals that allow you to create videos from a video or photo?

After the meteoric rise of generative artificial intelligence (GAI) models that transmute instructions into images, the new advance in this field are systems that generate clips from text. What Sora is like and what other tools are available for creating cinematic content.

From a technical point of view, according to OpenAI, what distinguishes Sora – which derives from the Japanese word “sky” – is the quality of its frames and also the ease of composing clips of up to a minute, longer than the fragments they offer. rivals.

Another amazing feature is that it doesn’t just produce short films that satisfy audience demands requiresbut does so in a way that demonstrates an emerging understanding of audiovisual narrative.

In turn, it includes a facial expression module, which allows you to animate figures with a high degree of emotion and gestures based on the input text, thus increasing the coherence of the images.

And although they are not required previous knowledge In handling cameras or directing, Sora poses no threat to Hollywood, as films cannot be made by cooking 60-second snippets. The disruption will happen on TikTok and Instagram, where the brief predominates.

The other side of the coin is its simplicity in becoming a quick and effective solution spread misinformationmaking it even more difficult to distinguish the real from the illusory.

The previews that the company has shared on its website indicate that a new creative frontier could spread in this area. The biological texture of a squirrel’s fur, the fear reflected in an astronaut’s gaze or the irregular step of a woman wandering through the crowd are some examples of his detailed setting.

Through X, the tech brand explained that it is located in a research phase. They are working with members of what is known as Red Teamcomposed of academic experts in the field of security, to address content that incites entertainment and glorifies some prejudices.

And although Sora is the prototype that generates the most expectations – due to its relationship with GPT and Dall-e – there are other models already operational, such as Stable Video Diffusion, Runway, Wonder Studio, which are a few meters ahead.

Stable video streaming

This AI engine designed to compose images from text has extended its roots to videos. It belongs to the company Stability AI and its distinctive feature is being of Open Sourceso that any developer can review its structure, make changes and create tools from the original.

The model runs through deep neural networks, trained with thousands of hours of video. From a textual description or an image it is possible to create sequences of 14 to 25 frames, with customizable speeds from 3 to 30 frames per second and a maximum duration of 5 seconds.

One of its technical possibilities is to simulate the effect of a camera moving around an object or person. The same trick is used to approximate a face. In both cases it is also possible to adjust the speed at which the lens moves.

And although all these solutions are in a phase of full evolution, with permanent innovations carrying a greater risk in the results, the stitches are still visible in this application.

However, since it is a product that can be used free For non-commercial purposes, this unique feature gives it an extra level of indulgence that others don’t have.

“We are working daily to optimize the model with new versions that improve video performance and stability,” emphasizes Tom Mason, CTO of Stability.

runway

It is the most established platform, as its Gen 2 technology (soon to be replaced by Gen 3) has managed to overcome the methodological inaccuracies faced by many of its competitors.

One of its attractions is intervening on an already recorded video. That is, if you have a shot of a man walking his dog, you can ask him to replace the animal with a lion or replace the urban background with a jungle background.

Another option is to animate a video without providing too many instructions. Once loaded, the software applies its alchemy, and once returned, you can be asked to apply new changes or undo what you’ve done.

Runway ML includes the ability to generate 3D models from images and descriptions. This allows designers to materialize their three-dimensional ideas without the need for modeling knowledge.

There is a plan basic and free which limits video length to 16 seconds and includes 3 projects per month. The standard subscription costs $12 and the Pro subscription goes up to $28. Outside of this menu, the cost to adopt Gen-2 is 0.05 cents per second, which translates to $3 per minute.

Study of Wonders

With a highly professional edge, the mission of this model suitable for independent productions is to simplify the film creation process through computer graphics (CGI), using artificial intelligence to make the process less cumbersome and more accessible.

Its developers claim that CGI no longer involves expensive motion capture equipment and that adding special effects It’s no more complex than applying filters in Photoshop.

With this platform you can easily produce films by selecting actors and assigning filler characters so that the system takes care of animating them and introducing them into each scene.

The advantage of Wonder Studio is its ability to resolve all the nuances of the production, from lighting, composition and camera movement, to the facial expression of the actors. With a single camera it is possible to obtain a complete and quality production.

To do this, it uses machine learning algorithms that draw a realistic 3D model of facial and body movements, allowing them to be coupled without being noticed in a video sequence.