# Lyria 3 and Lyria 3.5 Lyria is Google DeepMind's family of music models. Lyria 3 Clip makes 30-second clips and loops, while Lyria 3 Pro and Lyria 3.5 make full-length songs with verses, choruses and bridges. Each takes a text prompt and up to 10 images and returns an MP3 plus lyrics. In Sloosh they run in Spacelab's Music & Sound node. Web page: https://sloosh.ai/models/lyria [How to use Lyria 3 and Lyria 3.5](https://sloosh.ai/docs/models#audio-models) ## Key facts in Sloosh | | Lyria 3 Clip | Lyria 3 Pro | Lyria 3.5 | | --- | --- | --- | --- | | Made by | Google DeepMind | Google DeepMind | Google DeepMind | | Where | Spacelab, Music & Sound node (default model) | Spacelab, Music & Sound node | Spacelab, Music & Sound node | | Makes | 30-second music clip, MP3, plus lyrics | full-length song, MP3, plus lyrics | full-length song, 44.1 kHz stereo MP3, plus lyrics | | Inputs | text prompt; up to 10 reference images | text prompt; up to 10 reference images | text prompt; up to 10 reference images | | Length | fixed 30 seconds | no duration field; set length and structure in the prompt | no duration field; Google says "a couple of minutes", controllable in the prompt | | Lyrics input | no (write lyric lines in the prompt) | no | no | ## The models ### Lyria 3 Clip Google's short-clip model: "Optimized for generating short musical clips, loops, and previews up to 30 seconds" (Gemini API docs: Models). It always makes a 30-second clip, and Google suggests iterating on Clip before making a full song (Gemini API docs: Music generation). It is the model the Music & Sound node starts on. ### Lyria 3 Pro Launched with Lyria 3 Clip on 25 March 2026 for full-length songs (Gemini API changelog). Google's models page now lists it as the "previous generation" full-song model, after Lyria 3.5 (Gemini API docs: Models). It stays available in Sloosh for workflows already built on it. ### Lyria 3.5 Generally available in the Gemini API since 3 September 2026 (Gemini API changelog). Google calls it its "flagship music generation model, optimized for full-length songs with complex structural coherence" (Gemini API docs: Models). It writes lyrics in the language of the prompt (Gemini API docs: Music generation). ## Best for - Quick 30-second beds for social clips and ads: Lyria 3 Clip (Gemini API docs: Models). - Music that matches a visual: wire up to 10 product or scene images into the node (Sloosh catalog; Gemini API changelog says Lyria takes text and image inputs). - Finished full songs with vocals, including lyrics in other languages: Lyria 3.5 (Gemini API docs: Music generation). - Drafting an idea on Clip, then making the final on Lyria 3.5 (Gemini API docs: Music generation). ## How to use Lyria 3 and Lyria 3.5 in Sloosh 1. Open Spacelab ([sloosh.ai/spacelab](https://sloosh.ai/spacelab), signed in) and add a Music & Sound node from the add-node flyout (Audio). 2. In the node's Model picker, choose Lyria 3 Clip (the default), Lyria 3 Pro or Lyria 3.5. 3. Wire a Prompt or Text Input node into the prompt. Optionally wire up to 10 images into reference images. 4. Run the node. The exact price shows in Spacelab before you run. 5. Wire the audio into Merge audio + video to put it under a clip, or into a Download node. ## Lyria 3 and Lyria 3.5 prompting tips - Be specific: name instruments, BPM, key, mood and structure (Gemini API docs: Music generation). - Use section tags such as [Verse], [Chorus] and [Bridge] to shape the song (Gemini API docs: Music generation). - For background music, ask for an instrumental-only track, for example "Instrumental only, no vocals" (Gemini API docs: Music generation). - Write the prompt in the language you want the lyrics in (Gemini API docs: Music generation). - Iterate with Clip first, then make the full song (Gemini API docs: Music generation). ## Lyria 3 and Lyria 3.5 example prompts **Prompt 1** ```text Cinematic orchestral build with soaring strings and big timpani hits, 90 BPM, D minor, hopeful ending, instrumental only, no vocals. ``` **Prompt 2** ```text [Verse] soft acoustic guitar and a warm male vocal about a slow Sunday morning [Chorus] full band, bright and singable, 100 BPM, folk pop. ``` **Prompt 3** ```text Playful ukulele and whistling melody, bouncy and sunny, instrumental only, no vocals, for a kids' snack ad. ``` ## How it compares in Sloosh | Model | Length in Sloosh | Reference images | Lyrics input | | --- | --- | --- | --- | | Lyria 3 Clip | fixed 30 s | up to 10 | no | | Lyria 3 Pro | full song (length not settable) | up to 10 | no | | Lyria 3.5 | full song (length not settable) | up to 10 | no | | [ElevenLabs Music v2.5](https://sloosh.ai/models/elevenlabs-music-v2.5) | 3 to 60 s | no | no | | MiniMax Music 3 | 1 to 300 s | no | yes | From Sloosh's catalog rows. ## Where else it runs - Gemini API (Lyria 3 Clip, Lyria 3 Pro, Lyria 3.5) (Gemini API docs: Models, Oct 2026). ## Frequently asked questions ### What is Lyria 3? Lyria 3 is Google DeepMind's music model family launched on 25 March 2026, with Lyria 3 Clip for 30-second clips and Lyria 3 Pro for full songs (Gemini API changelog). Lyria 3.5 followed in September 2026. ### What is the difference between Lyria 3 Pro and Lyria 3.5? Lyria 3.5 is the newer one. Google lists it as its flagship full-song model and Lyria 3 Pro as the previous generation (Gemini API docs: Models). Both are in Sloosh. ### How long is a Lyria track? Lyria 3 Clip is always 30 seconds. Lyria 3.5 makes songs of "a couple of minutes", which you can steer in the prompt (Gemini API docs: Music generation); Sloosh's node has no duration field for the full-song models. ### Does Lyria write lyrics? Yes. The response includes the generated lyrics with the audio (Gemini API docs: Music generation), and Sloosh returns the MP3 plus lyrics. ### Can Lyria make instrumental music? Yes. Ask for an instrumental-only track in the prompt (Gemini API docs: Music generation). ### Can Lyria use images in Sloosh? Yes. Wire up to 10 images into the Music & Sound node's reference images. ### Does Lyria music carry a watermark? Yes. Google says all generated audio includes an imperceptible SynthID audio watermark (Gemini API docs: Music generation). ### How do I add Lyria music to a video in Sloosh? Wire the track and the video into Merge audio + video. The video keeps its length and longer audio is cut off. ## Related - Models: [ElevenLabs Music v2.5](https://sloosh.ai/models/elevenlabs-music-v2.5), MiniMax Music 3, [ElevenLabs Sound Effects v2](https://sloosh.ai/models/elevenlabs-sound-effects-v2) - Spaces: [Add a soundtrack](https://sloosh.ai/spaces/add-a-soundtrack) - Docs: [Models](https://sloosh.ai/docs/models.md#audio-models), [Edit nodes](https://sloosh.ai/docs/nodes-edit.md#merge-audio-video) ## Sources - [Sloosh docs: Models (Audio models table)](https://sloosh.ai/docs/models.md) - [Sloosh Space: Add a Soundtrack](https://sloosh.ai/spaces/add-a-soundtrack/agents.md) - [Sloosh docs: Edit nodes](https://sloosh.ai/docs/nodes-edit.md) - [Gemini API docs: Music generation](https://ai.google.dev/gemini-api/docs/music-generation) - [Gemini API docs: Models](https://ai.google.dev/gemini-api/docs/models) - [Gemini API docs: Changelog](https://ai.google.dev/gemini-api/docs/changelog) - [Google DeepMind: Lyria](https://deepmind.google/models/lyria/)