Working with Lyria 3
Two very different capabilities share this family. Lyria 3 Pro generates full-length songs from text or image prompts and returns high-fidelity MP3 with vocals, lyrics and song structure; Lyria 3 Clip does the same in 30-second form for fast drafts. Gemini 2.5 Flash is the low-latency multimodal model that accepts text, image, audio, video and document input.
When to choose Lyria 3
Come here for AI music with lyrics and vocals — no other family in the catalog writes complete songs. For image or video generation from Google, the Veo, Imagen and Nano Banana families are the right destinations.
Choosing between the variants
Lyria 3 Pro is for a finished full-length track; Lyria 3 Clip returns a 30-second draft far faster and cheaper, which is the sensible way to test a musical direction. Gemini 2.5 Flash is the text and reasoning model and produces no audio at all.
About Lyria 3
- Does Lyria 3 generate vocals and lyrics?
- Yes. Lyria 3 Pro returns high-fidelity MP3 with vocals, lyrics and a song structure rather than an instrumental bed.
- Can Lyria 3 take an image as the prompt?
- It accepts text or image prompts, so a mood board or a single picture can drive the musical direction of the track.
- What is Gemini 2.5 Flash doing in a music family?
- It is the multimodal text model shipped under the same Gemini family slug — fast, lightweight, and able to read text, image, audio, video and documents.