Google’s Deep mind The artificial intelligence laboratory is preparing to solve the problem of creating background music and even dialogue, which are among the biggest shortcomings of the recently popular artificial intelligence for video creation. V2A The lab shares its advances in what it calls (beyond video audio) technology and can also be combined with video creation tools such as Google Veo and OpenAI Sora.

According to the blog post shared by the DeepMind team, the system can understand raw pixels. and stated that he could combine this information with text prompts. Sound effects are also created during this competition. This tool can also be used for silent movies or other videos without sound.

The DeepMind team is confident in its tools

The first artificial intelligence that can actually create sound Deep mind It’s not the technology he developed, and it won’t be the last. ElevenLabs had released this kind of tool before. On the other hand, the DeepMind team is working on its own tools. “It differs from existing video-based audio solutions in that it can both understand pixels and add text prompts, which is optional.” He uses the expression.

DeepMind trained researchers in their work to develop this technology with data consisting of videos, sounds, detailed audio descriptions and transcripts produced by artificial intelligence. Researchers too V2A technologyThey also state that they are working on existing issues, such as a decrease in the sound quality of the output if there is distortion in the source video. The DeepMind team also promised to be “subject to rigorous safety evaluation and testing” before making the technology available to the public.