OpenAI's Next Music Generator Might Be for Sora

According to US sources, the AI giant is working on a new tool for artificial music. This could also be used for scoring videos.

listen Print view
Symbolic image AI music: silhouette of a human head, with artificial representation of music-symbolizing waves and bar graphs

(Image: whiteMocca/Shutterstock.com)

3 min. read

While OpenAI is considered a technology leader in chatbots and video generators, this is not the case for fully AI-generated music – but that is set to change soon. At least, if a report by "The Information", based on several unnamed sources, is to be believed. OpenAI is reportedly working on a new music generator, following the company's development of the Musenet and Jukebox models. However, both are not publicly accessible.

The new tool, which does not yet have a name, is said to work similarly to the video generator Sora 2. According to the report, music can be created with both text and audio prompts. Sora 2 can also handle inputs via text, image, or video. An existing vocal recording, to which the AI can invent a guitar accompaniment, is mentioned as an example for the music generator.

Videos by heise

However, the system is also said to be able to create complete pieces, including vocals from scratch. This is a capability that ChatGPT does not yet offer. Other AI providers like Suno or Udio can do exactly that, however. With these services, it is also possible to calculate music in the style of well-known genres. The results appear more or less authentic depending on the prompt effort.

How OpenAI intends to offer its new music AI is not yet clear from the report. Sora 2 appeared as a standalone app. And Sora could also be where the greatest benefit lies. While the current version of the video generator can invent dialogues, sound effects, and simple musical snippets, it cannot create a complete soundtrack. In the human-made film business, musicians often compose specifically for individual scenes based on the finished work, but editors or directors do not always adopt this.

Empfohlener redaktioneller Inhalt

Mit Ihrer Zustimmung wird hier ein externes YouTube-Video (Google Ireland Limited) geladen.

Ich bin damit einverstanden, dass mir externe Inhalte angezeigt werden. Damit können personenbezogene Daten an Drittplattformen (Google Ireland Limited) übermittelt werden. Mehr dazu in unserer Datenschutzerklärung.

What Sora 2 can do based on the context of a scene is also to invent the content for it, as we have already tried out ourselves. If OpenAI succeeds in creating music that truly matches the content and mood of a video, it would be a new dimension. "The Information" suspects that OpenAI is initially targeting the advertising clip market with the combination of music and video. Adobe attempted something similar with its image generator and the slogan "Skip the Photoshoot" a year and a half ago and faced strong backlash from the advertising industry.

(nie)

Don't miss any news – follow us on Facebook, LinkedIn or Mastodon.

This article was originally published in German. It was translated with technical assistance and editorially reviewed before publication.