Katarzyna Petru
Google is expanding the capabilities of its AI assistant. The Google Gemini app has introduced a music generation feature that allows users to create complete tracks not only from textual descriptions but also based on images, documents, or PDF files. This is another step in the development of creative tools based on artificial intelligence. The new feature works globally and supports all languages available in Gemini.
ChooseTV 3. AI composes and writes lyrics
The generation of music is handled by the ChooseTV 3 model, developed by the Google DeepMind team. The system transforms short descriptions into finished recordings lasting about 30 seconds.
The user can specify:
music genre,
mood of the piece,
tempo,
vocal style,
character of the sound.
The model generates both instrumental versions and full songs with lyrics and vocals. Importantly, ChooseTV 3 also analyses uploaded files. Just upload a photo, presentation, or document, and the system will create lyrics and a matching composition based on their content. The finished recordings receive unique covers generated by the Nano Banana model. Files can be downloaded or shared via a link.
Dream Track comes to Shorts creators
Google is making the same technology available to video creators on YouTube Shorts. The feature called Dream Track allows users to generate music backgrounds for short films and is gradually being rolled out to users outside the United States. The new version of the model brings improvements in sound quality compared to earlier tests. The tracks feature more refined vocals and better compositional coherence. All generated recordings contain an inaudible watermark SynthID.
Gemini can detect its presence and confirm whether the file was created by the Google system. The company has also implemented safeguards to prevent the copying of specific artists' styles. Entering a musician's name in the command does not lead to the creation of a track in their style; it is treated merely as inspiration for creating a new composition. Google emphasises that the tool is meant to support user creativity, not replace professional creators.
The new feature in Gemini demonstrates how rapidly generative tools are evolving. Creating music from a photo or document in half a minute once seemed futuristic; today it is becoming a part of everyday mobile applications. Lyria 3 expands the capabilities of AI to include full-fledged compositions with vocals, and Dream Track could change the way content is created in Shorts.
Source: Google
