Joerg Hiller
Feb 23, 2026 17:52
Google’s Lyria 3 AI music mannequin now creates 30-second tracks from textual content or pictures in Gemini app, that includes computerized lyrics and SynthID watermarking.
Google DeepMind has rolled out Lyria 3, its most superior generative music mannequin, throughout the Gemini app as of February 18, 2026. The software lets customers create authentic 30-second tracks utilizing textual content prompts, images, or video uploads—a notable growth of AI capabilities into inventive audio manufacturing.
The improve from Lyria 2 brings computerized lyrics technology, which means customers now not want to put in writing their very own phrases. Merely describe a theme like “a love tune” or “a monitor about success,” and the mannequin handles the remaining. For many who desire management, customized lyrics nonetheless work—simply prefix them with “Lyrics:” in your immediate.
How It Works
Entry Lyria 3 via the “create music” possibility within the Gemini app. From there, you have received a number of enter strategies. Textual content prompts can specify style, period, devices, vocal model, and tune dynamics. Need ’90s skate punk with quick drums? Executed. Desire mixing Okay-pop with Motown influences? The mannequin accommodates style mashups.
Visible inputs add one other dimension. Add a trip photograph, pet image, or authentic paintings, and Lyria analyzes parts like topics, clothes, and backgrounds to generate a musical match. The identical applies to video clips.
Artistic Controls
Specificity pays off with Lyria 3. Customers can outline whether or not vocals come from a male or feminine singer, specify vocal vary from baritone to soprano, and add texture descriptors like “gravelly” or “breathy.” Instrumental decisions prolong past style defaults—including ’80s synth to a Fifties jazz basis is honest recreation.
Track dynamics additionally reply to prompting. A quiet piano intro constructing to an explosive refrain, purely instrumental sections, or vocals that cut up into harmonies as tracks progress—all achievable via detailed directions.
Safeguards Constructed In
Google has carried out SynthID watermarking throughout all Lyria 3 outputs, making AI-generated content material imperceptibly traceable. The corporate additionally filters artist identify references in prompts, treating them as broad stylistic inspiration quite than targets for direct imitation—a transfer addressing ongoing issues about AI music mimicking particular performers.
Completed tracks obtain as MP3 or MP4 information for sharing on social platforms or messaging apps. The 30-second restrict retains outputs centered, although it additionally constrains extra bold musical concepts.
Google mother or father Alphabet trades at $313.99 as of February 23, with market cap sitting at $3.80 trillion. Whether or not AI music technology strikes that needle stays unclear, however the function positions Gemini as a extra versatile inventive assistant in opposition to competing AI platforms nonetheless centered totally on textual content and picture technology.
Picture supply: Shutterstock
