lyria 3 5

🎨 lyria
← Back to lyria 📄 Raw Markdown

Lyria 3.5 (lyria-3.5) Model Card #

Overview #

Lyria 3.5 (lyria-3.5) is Google's state-of-the-art flagship music generation model. It produces high-fidelity 44.1 kHz stereo audio, full songs with structural arrangements (intro, verses, chorus, bridge, solo, outro), expressive vocal synthesis, and dynamic intensity shaping from text prompts or visual reference images.

Technical Specifications #

Property Value
Input Modalities Text, Image (up to 10 reference images)
Output Modalities Audio (MP3 / WAV) and Text (Lyrics / Structure)
Input Token Limit 131,072 tokens
Audio Output Duration Controllable multi-minute songs (up to ~3 minutes)
Audio Quality 44.1 kHz stereo, 192 kbps bitrate
Supported MIME Types audio/mp3, audio/wav
Supported Languages English, German, Spanish, French, Hindi, Japanese, Korean, Portuguese
Structural Controls Timestamps, BPM, Key, Section Tags, Dynamic Intensity
Watermarking SynthID (Audio) + C2PA Content Credentials

Key Features & Capabilities #

1. Structural Arrangement & Timestamp Control #

Lyria 3.5 interprets structured prompt timelines to guide instrument entrances, transitions, vocal delivery, and dynamic swells:

text 
[0:00 - 0:15] Intro: Atmospheric Rhodes electric piano with ambient vinyl crackle in D minor.
[0:15 - 0:45] Verse 1: Warm bassline and boom-bap drums enter with expressive vocals.
[0:45 - 1:15] Chorus: Full melodic progression with saxophone countermelodies and rich harmonies.
[1:15 - 1:35] Bridge: Stripped-back percussion and solo piano building tension.
[1:35 - 2:05] Final Chorus: Maximum dynamic intensity with full brass accents and rhythm section.
[2:05 - 2:20] Outro: Slow guitar strumming gently fading to silence.

2. User-Defined Lyrics & Section Tags #

Provide full song lyrics structured by section:

text 
Compose an uplifting indie rock anthem at 128 BPM in G Major:

[Verse 1]
City skyline shining in the dusk
Running free on roads of copper rust

[Chorus]
We ignite the neon spark tonight
Chasing dreams until the morning light

3. Multimodal Image-to-Music Guidance #

Supply up to 10 reference images (concept art, photos, storyboards) to direct the emotional palette, genre mood, and aesthetic tone of the composition.

4. Clean Instrumental Backing Tracks #

Add "Instrumental only, no vocals" to synthesize pristine background beds, game soundtracks, or podcast scoring without vocal bleed or artifacts.

SDK Code Examples #

1. Full Song with Timestamps & Lyrics (Python - Interactions API) #

python 
import base64
from google import genai

client = genai.Client()

prompt = """
An atmospheric lo-fi jazz track at 85 BPM in E Minor:

[0:00 - 0:20] Intro: Warm upright bass and gentle brush drums.
[0:20 - 0:50] Verse 1: Mellow electric piano melody with subtle trumpet accents.
[0:50 - 1:20] Chorus: Full ensemble with rich saxophone harmonies.
[1:20 - 1:40] Outro: Slow piano chords fading into rain sounds.
Instrumental only, no vocals.
"""

interaction = client.interactions.create(
    model="lyria-3.5",
    input=prompt,
    response_format={"type": "audio"},
)

if interaction.output_audio:
    with open("lofi_track.mp3", "wb") as f:
        f.write(base64.b64decode(interaction.output_audio.data))

if interaction.output_text:
    with open("lofi_track_structure.txt", "w") as f:
        f.write(interaction.output_text)

2. Multimodal Concept-to-Soundtrack (Python) #

python 
import base64
from google import genai

client = genai.Client()

def read_b64(path):
    with open(path, "rb") as f:
        return base64.b64encode(f.read()).decode("utf-8")

interaction = client.interactions.create(
    model="lyria-3.5",
    input=[
        {"type": "image", "data": read_b64("concept_art.jpg"), "mime_type": "image/jpeg"},
        {
            "type": "text",
            "text": "Compose a sweeping cinematic orchestral film score reflecting the epic scale and mysterious tone of this concept art. Instrumental only.",
        },
    ],
)

if interaction.output_audio:
    with open("film_score.mp3", "wb") as f:
        f.write(base64.b64decode(interaction.output_audio.data))

3. JavaScript / TypeScript SDK (@google/genai) #

typescript 
import { GoogleGenAI } from "@google/genai";
import * as fs from "node:fs";

const ai = new GoogleGenAI({});

async function generateSong() {
  const interaction = await ai.interactions.create({
    model: "lyria-3.5",
    input: "An uplifting synthwave anthem at 124 BPM in A Minor with retro analog synthesizers and punchy drum machine. Instrumental only.",
  });

  if (interaction.output_audio) {
    fs.writeFileSync("synthwave_anthem.mp3", Buffer.from(interaction.output_audio.data, "base64"));
  }
}

generateSong();