Long before the advent of generative artificial intelligence sing songs for photos, animations were made. Microsoft’s artificial intelligence tool now turns photos into more realistic videos. Not only that, the tool can also make photos sing whatever they want, in any way they want: this goes for songs too.

Introduced by Microsoft Asia Research and VASA-1 The so-called artificial intelligence tool can take any photo or drawing and combine it with an existing audio file. The new artificial intelligence tool can create facial expressions and head movements. It also produces mouth movements that match speech.

It is ‘for now’ clear that the images are the work of artificial intelligence

Although in the images taken by VASA-1 mouth and head movements Although it looks a bit robotic, if we look closely there are shifts in voice and lip sync. However, it also happens that these technologies can be used to create fake images or deepfake videos over time. Researchers are also aware of this situation and have therefore not shared a useful demo or API. This technology tooresponsibleHe stated that they wanted to make sure it would be used.

Still, researchers believe this technology can be used for good purposes. VoxCeleb2 It is stated that communication with artificial intelligence can be strengthened, new tools can be developed in the field of education and communication problems can be solved, thanks to the artificial intelligence tool trained with the images of 6112 celebrities with the so-called dataset.

You can access the research and demo images published by Microsoft here.