Google brings Lyria 3.5 to Gemini: begin with the job the music needs to do

Google announced Lyria 3.5 in Gemini on September 4. The announcement describes vocal or instrumental choices, templates, short or longer tracks, and availability through the Gemini app and developer surfaces including the Gemini API. It does not establish that every surface exposes identical controls.

Turn a template into a brief

A template can get you started, but a use case still needs decisions. For a video opening, write down when the first spoken sentence begins, whether the opening needs a strong beat, and where the music must leave space. These constraints are more useful than a long string of mood adjectives.

Try an original brief such as: “An instrumental opening for a quiet walking video. Soft piano is the foreground; a warm pad enters later. The energy stays even, with room for narration.” The example is a writing exercise, not a tested formula for a specific model.

After generating, place the result under the actual video. Check the first edit, the first sentence and the final transition. Save a note for each location instead of giving the whole piece a single score. You will then know whether to change the opening, the density or the ending on the next attempt.

Keep app and API decisions separate

The announcement is a discovery source. If an agent will call the model, follow the current developer documentation for model names, input fields and output handling. Do not copy interface labels into an API request by guesswork.

You can keep the creative brief and musical choices in MusicPrompt Studio even when a different tool renders the audio. The learning path explains how to turn those choices into repeatable exercises. Compare tools using the same musical job, not unrelated demonstration tracks.