- API Pricing and Enterprise Integration Expand Professional Access
- Multimodal Inputs and Expanded Product Ecosystem
The release targets content creators, developers, and enterprises seeking royalty-free music for videos, podcasts, and games. Lyria 3 Pro is available immediately to Google AI Pro and Ultra subscribers through the Gemini app, with enterprise access via Vertex AI public preview and developer integration through Google AI Studio and the Gemini API priced at $0.08 per generation.
API Pricing and Enterprise Integration Expand Professional Access
Google introduced tiered API pricing for the Lyria 3 family, charging $0.08 per track for the full-length Lyria 3 Pro model and $0.04 for the 30-second Lyria 3 Clip variant optimized for rapid prototyping. Developers can integrate the music generation into custom applications via the Gemini API, while enterprises gain access through Vertex AI for scalable soundtrack production in gaming and media workflows.
The model supports 48kHz stereo audio output and features granular controls including tempo conditioning and time-aligned lyrics. All generated tracks carry SynthID watermarks, imperceptible markers designed to survive compression and editing allowing platforms to identify AI-generated content regardless of subsequent modifications.
Multimodal Inputs and Expanded Product Ecosystem
Lyria 3 Pro introduces image-to-music capabilities alongside text prompts, enabling users to upload images that influence the mood, style, and atmosphere of generated tracks. Google is embedding the model across its product ecosystem including Google Vids for video soundtracks, ProducerAI for collaborative music creation, and the Gemini app where paid subscribers receive daily generation limits ranging from 10 tracks for AI Plus users to 50 for Ultra tier.
The system supports eight languages at launch: English, Japanese, Korean, Hindi, Spanish, Portuguese, German, and French and offers real-time generation via the Live API for interactive applications. Google emphasized the model was trained exclusively on data "that YouTube and Google has a right to use", explicitly stating it does not mimic specific artists but instead takes "broad inspiration" from stylistic references when prompted.
Source: https://blog.google/innovation-and-ai/technology/developers-tools/lyria-3-developers/