Google recently announced the “Lumiere” system, which works on the basis of generative artificial intelligence and allows video production by typing directly.
The new system, which you can watch the promotional video prepared below. Google Lumièreprepared on the basis of generative artificial intelligence and It can create five-second videos from what is written. The system, which can create higher quality videos than many similar ones, currently outputs 1024×1024 pixels (defined as low resolution by the company) and can customize the result as desired. Many examples shared by Google on this subject on the Github page here It is not yet known when or how Lumiere, on which work will continue, will reach the end user. Company Lumière The text-to-video model called “T2V”, which it uses on its basis, complete with text titles. He states that they trained using a dataset containing 30 million videos.
Google This is not the first time he has taken a step in this regard. The company had previously made a splash with “VideoPOET”, which it introduced with a focus on video production. VideoPoet, to remember again, is the company’s new LLM, that is, the big language model. This system, which is still under development, It can create 10-second videos from the entered sentences. And Here it performs better than many similar systems. Of course, the infrastructure you can see in the video below is not at a perfect level yet, but already reveals great potential for the future. Ability to animate not only text but also entered photos/images by looking at the given command. VideoPoetSince it is the new trend, it is specifically designed to produce content in vertical format. According to the information provided by Google researchers, artificial intelligence, which is not yet an option that can be used by everyone, such as DALL-E signed by OpenAI, In the future, it will also work in areas such as text-to-sound and audio-to-video.