Google Gemini: AI bot now generates complete songs including lyrics

After text, image and video, Google Gemini is now conquering the world of audio.proInduction. With the latest update, Google integrates the powerful Lyria 3 music model directly into its chatbot. Users can now create 30-second tracks that include not only the melody but also suitable vocals and lyrics.

Lyria 3: From text to finished production

Until now, Google's music experiments like MusicFX or Dream Track were often limited to short, instrumental clips. Lyria 3 goes a significant step further: The model understands context and mood so well that it combines complex, multi-layered instrumentals with human-sounding vocals. The multimodality is particularly interesting. Users don't necessarily have to describe how the song should sound. Uploading a photo or video is enough. Gemini analyzes the visual mood and composes a suitable soundtrack based on it—for example, a relaxed acoustic pop song for a hiking photo or an energetic beat for a sports video.

Deep editing and creative control

Compared to competitors like Suno or Udio, Gemini offers extensive customization options directly within the chat. After generating a track, users can modify details via a follow-up prompt. The system allows users to vary the tempo, adjust the musical style, or instruct the AI ​​to change the vocal style. Lyrics can also be modified or completely rewritten based on the prompt. Users with a specific inspiration in mind can also upload existing music files so that Gemini can use that style as a reference for the new composition.

Technical details and protection against misuse

The core of the function is Google DeepMind's Lyria 3 model, which can process tracks with a maximum length of 30 seconds. proEach song is automatically paired with a matching album cover created by the Nano Banana AI. To prevent copyright infringement claims, all tracks include an invisible SynthID watermark. Google also emphasizes that the model was trained not to copy existing artists. Output is automatically checked against databases to prevent plagiarism, and users have the option to report suspicious content directly.

availability and costs

The feature is being rolled out today for the Gemini web interface and mobile apps on Android and iOS. It's launching as a beta in several languages, including German, English, French, and Japanese. Music generation is available to both free users and paying subscribers, with the number of available credits depending on the subscription level. Users with an AI Premium plan receive correspondingly higher credit allowances for their productions.