Artificial intelligence madness: from now on Gemini creates music for you

According to the announcement, users can write Gemini a description of a song or instrumental piece, get an immediate result, and then refine the instructions until a version closer to what they envisioned is obtained. Google presents it as an approach that lowers barriers and enables quick creation even for those who have no musical knowledge. The examples presented in the demos sound convincing in terms of the instruments and the sound, but even there the familiar weak point of generative music is hinted at, words that sometimes sound strange, kitschy or unnatural, even if the arrangement itself is polished.

One of the significant innovations is the transition from a work that is based only on text to a work that is also influenced by visual materials. In Gemini you can, according to Google, upload a photo or video and ask for music that fits the atmosphere, rhythm or nature of the scene. In the meantime, Google notes that the pieces can come together with an album cover that is produced using an image model called Nano Banana, which is aimed in advance at the uses of sharing on networks and short video platforms.

Google also connects the new ability to YouTube, through a feature called Dream Track, which is intended for creators of shorts and produces backing tracks. This means that creating music does not stay within the confines of the chatbot, but goes directly into the pipeline where millions of short videos are created every day. For YouTube, this is another tool that reduces dependence on existing music libraries and usage rights, and on the other hand may deepen the old debate about originality, ownership and compensation in the age of production at the click of a button.

At the center of the move is also the issue of trust. Google says that every segment produced in the system is imperceptibly marked using SynthID, a marking technology designed to identify content created by Google’s artificial intelligence. The company also promoted last year a test tool called SynthID Detector designed to help locate such markings in files, as part of a broader effort to combat impersonation and misleading content. The limitation here is clear, the marking and testing will work mainly for content created with Google’s tools, and not necessarily for products that come from other companies.

So is Google really threatening Suno, one of the most popular AI music creation tools? The answer is complex. On the one hand, Suno already offers the creation of longer songs today, according to information that appears in its help database, advanced generations are able to create many minutes and even more before extension, while Google currently limits Gemini to 30 seconds. On the other hand, Google’s strength lies elsewhere, in distribution. When the ability sits inside Gemini, inside Google products, and with a connection to YouTube, it can reach users who won’t bother signing up for a dedicated service. The scenario where music is created directly from a conversation, and then integrated into messages or other Google applications, is already hovering in the background even if the company has not officially announced it as part of the launch.

According to Google, the ability begins to open today to users aged 18 and over, and in the first phase it is available in a specified number of languages, English, Spanish, German, French, Hindi, Japanese, Korean and Portuguese. For an Israeli audience, this means that creation in Hebrew is not yet part of the initial launch, but historically, when such capabilities enter Google’s core products, they tend to expand gradually. The big question is not only when it will reach Hebrew, but what the market will look like when the produced music becomes another standard feature within the applications we all already use.