A short film of a repaired chair does not necessarily need a song. It may need a quiet opening while someone explains the damage, a little movement while the work progresses, and a sense of arrival when the finished object appears. That is a creative job a music prompt can describe more precisely than “make something uplifting”.
Google added Lyria 3.5 to the Gemini app on 4 September 2026, with vocal or instrumental styles, genre choices and short or longer tracks. Lyria is the music-generation model; its developer documentation describes audio generation from text and image inputs. The workflow here uses the Gemini app rather than the API.[1][3]
Prepare the brief before opening the generator
Use a personal video you control and choose the section that needs music. For this exercise, imagine a 30-second edit: an opening view of the worn chair, a few moments of repair and a final reveal. The timings and prompt below are our invented creative example, not a soundtrack generated or auditioned for this article.
Decide what the audience must hear. If there is spoken narration, a busy melody or prominent vocal could compete with it. If the picture already contains a strong action rhythm, the music may only need to support that rhythm rather than add another. These are editing choices; the generator cannot infer them reliably from a genre label alone.
Create an instrumental draft
Google’s help page requires an eligible signed-in account, age 18 or over, and Keep Activity enabled. In the web app, open Upload & tools and choose Create music. The documented short option is about one minute; a full track is about two to three minutes and requires the Pro model. For a 30-second film, generate a short draft and select a usable passage in the video editor.[2]
The brief has four jobs: it identifies the setting, selects a restrained palette, describes how the music should develop, and names the ending. The model may not place each change exactly where requested. Treat the first output as a draft to fit against the actual edit rather than assuming that a detailed prompt creates frame-accurate scoring.
Listen against the picture, not in isolation
Place a roughly 30-second passage against the picture, keeping the original draft so you can try a different section. Listen first for clear narration, then for whether the lift supports the reveal. Adjust the music’s position or trim at a musical phrase boundary. If the selected passage cannot end naturally, try a different section or regenerate with a simpler ending.
Change one problem at a time. If the opening is too busy, ask for less percussion and a simpler guitar pattern there. If the ending feels abrupt, ask for a more settled final phrase. A request such as “make it better” discards the useful diagnosis and gives the next generation several unrelated ways to change.
This is also a reason to preserve the previous draft. A new version can solve one problem while losing a quality you liked. Comparing two specific passages—rather than repeatedly replacing the whole piece on first impression—makes the process more deliberate.
Download the file that suits the edit
Google documents an audio-only MP3 download and an MP4 version with cover art. Choose MP3 when the video editor needs just the music. The help page also describes an embedded digital watermark identifying AI-generated music; a watermark is not a guarantee about ownership or every possible reuse.[2]
When the feature is missing, check the account’s eligibility and whether the web version exposes Create music before spending time rewriting prompts. The steps here are based on Google’s documentation, not a hands-on trial, and interface details can differ from the help page retrieved for this guide.[2]
A useful result is not simply a track that sounds impressive on its own. It is a track that leaves space for the voice, supports the change in the picture and arrives at an ending with it. The creative brief improves when those jobs become explicit.
Sources & further reading
- Create your best tracks yet with Lyria 3.5 in Gemini Google · 4 Sep 2026 · Accessed 14 Oct 2026
- Generate music with Gemini Apps Google Gemini Help · Accessed 15 Oct 2026
- Lyria 3.5 Google AI for Developers · Accessed 14 Oct 2026