Google Gemini now creates music from text, images, and videos
Users can now create AI music with Gemini. The tracks are generated based on text instructions, images, or videos. Google aims to respect copyrights.
(Image: whiteMocca/Shutterstock.com)
Google's AI model Gemini now allows users to create music using generative artificial intelligence (AGI). The 30-second tracks are generated based on user text prompts, but the AI can also compose music based on images or videos, including vocals. The produced music pieces are embedded with an invisible watermark to identify them as AI-generated. Google promises to comply with copyright laws.
For the production of AI music, the company uses Google DeepMind's new generative music model "Lyria", now in its third version. Google has been working on it since 2023 and announced it at Google I/O in 2024 alongside other AI models. Nevertheless, according to Google, the function in Gemini is still in beta status. It is available to all adult Gemini users in English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese.
Texts, images, and videos for inspiration
Music generation is done via text prompts, with no lyrics required. Lyria 3 can compose the songs itself. Users should specify a music genre, but details about vocals, pitch, or tempo are optional, Google writes. An example from the company: "Create a fast (120 BPM) soul-funk track with a warm, female soprano voice." However, imprecise prompts are also possible and result in an acceptable track, for example: "Trance music from the 90s that starts slow but feels uplifting."
(Image:Â Google)
Additionally, Gemini can use images and videos for music generation if the user uploads them to the AI model. This is intended to inspire Lyria 3 and set a mood for the AI: "Use these photos to create a track about my dog on a hike in the forest." Gemini also automatically creates a graphic as cover art to match the music piece. This is produced by the AI image generator Nano Banana, also integrated into Gemini.
YouTube background music with watermark
Google is now also using Lyria 3 within YouTube. Content creators can add suitable background music to their "Shorts" using it. This feature, called "Dream Track", was previously only available as an experiment in the USA but can now be used in other countries as well. This is part of Google's strategy to support content creators on YouTube with AI.
Videos by heise
While AI-generated images and videos can be visually marked to indicate they are AI-generated, this remains hidden for music. However, the company promises that tracks created by Lyria 3 in both Gemini and YouTube will be watermarked with Google's watermark SynthID for AI content. Although this is not audible, it can be verified with the help of Gemini.
Copyrights likely respected
Google emphasizes that the generation of AI music does not infringe on copyrights and is not intended to imitate artists. Existing music pieces are merely used as inspiration. "Our training for Lyria 3 is designed to use music that YouTube and Google are authorized to use according to our terms of service, partner agreements, and applicable law," Google explained to Bloomberg. Before Gemini outputs music, filters are also supposed to "check the tracks against existing content."
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.
The generation of AI music is also available to Gemini users without a paid subscription. In this case, Google limits this to 10 tracks per day. Subscribers of Google AI Plus, Pro, or Ultra, on the other hand, can have 20, 50, or even 100 music pieces created per day.
(fds)
