SuperAPI AI Directory
Partner with us中文AI newsFavorites
DirectoryNews and tutorials

Lyria 3 - Google DeepMind's new-generation AI music generation model

Author: AI小集 · Original source: AI工具集

What is Lyria 3

Lyria 3 is the latest-generation AI music generationmodel launched by Google DeepMind, and it has now been integrated into the Gemini app. Compared with previous generations, Lyria 3 has achieved a major breakthrough: users no longer need to write lyrics themselves. With just a text description or by uploading a photo or video, they can generate a 30-second high-quality music clip with automatic lyric writing, composition, and vocals in one click. The system supports fine-grained style control, covering elements such as rhythm, mood, and vocals, and automatically generates accompanying cover artwork for each piece. Lyria 3 has built-in SynthID watermarking technology to track and verify AI-generated content, and it also includes copyright protection mechanisms to avoid directly imitating the works of existing artists. The model currently supports eight languages: English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese. It is free for Gemini users aged 18 and above, and also provides AI soundtrack capabilities for YouTube Dream Track, making it suitable for scenarios such as short video creation, personal entertainment, and creative expression.

Lyria 3

Lyria 3's main features

  • Multimodal music generation: Supports generating 30-second high-quality music clips through text descriptions or by uploading photos or videos, with AI automatically matching mood and style.
  • Automatic lyric creation: No need for users to provide lyrics; the system automatically generates complete lyrics and vocal performance based on the prompt.
  • Fine-grained style control: Music style, vocal performance, tempo, and other elements can be adjusted to achieve more realistic and complex musical arrangements.
  • Intelligent cover generation: Each song has accompanying cover artwork automatically generated by Nano Banana AI.
  • Multilingual support: Currently supports eight languages: English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese.
  • Copyright safety protection: Built-in filters avoid generating content that is too similar to existing works; mentions of specific artists are used only as style inspiration rather than direct copying.
  • SynthID watermark tracking: All generated music is embedded with an imperceptible digital watermark, and uploaded audio can be verified to determine whether it was generated by Google AI.
  • Multi-platform integration: Already integrated into the Gemini app and YouTube Dream Track, supporting downloads in MP3 audio or MP4 video format.

Lyria 3's technical principles

  • Multimodal understanding architecture: Can process text, image, and video inputs simultaneously, parsing content mood and scenes through a vision-language model and converting them into music generation instructions.
  • End-to-end music generation: Uses a unified neural network architecture to integrate lyric generation, melody creation, arrangement, and vocal synthesis into a single process rather than handling them in stages.
  • SynthID audio watermarking: Embeds a digital fingerprint imperceptible to the human ear during the generation process, using frequency-domain transformation technology to hide identifying information in the audio waveform, supporting subsequent traceability verification.
  • Copyright protection mechanism: Based on a large-scale audio fingerprint database and similarity detection algorithms, it compares generated content with existing copyrighted works in real time and triggers filtering or adjustment mechanisms.
  • Style control and constrained generation: Through conditional generation technology, user-specified parameters such as style, rhythm, and mood are injected into the generation process as constraints to ensure the output meets expectations.
  • Nano Banana Visual Generation: Integrates an image generation model to automatically generate matching cover artwork based on music style and lyrics theme.

How to Use Lyria 3

  • Access Point: Open the Gemini app (web or mobile), find and tap the "Music" option in the bottom toolbar.
  • Choose Generation Method: Supports three input methods—directly entering a text description, uploading a local photo, or uploading a video file.
  • Write a Prompt: Use text to describe the desired music style, mood, themed scene, etc., for example, "a nostalgic Afrobeat song about childhood memories."
  • Wait for Generation: After submission, in about 10-60 seconds, the system automatically completes lyric writing, composition, arrangement, and vocal synthesis, generating a 30-second music clip.
  • Preview and Adjust: Play the generated music; if unsatisfied, you can modify the prompt or adjust style, tempo, and other parameters to regenerate.
  • Download and Save: Supports downloading the work as MP3 audio format or MP4 video format with cover art, making it easy to share on social platforms.
  • Watermark Verification: If you need to verify whether a piece of audio is AI-generated, you can upload it to Gemini for SynthID watermark detection.
  • Usage Restrictions: Must be at least 18 years old; free users have a limit on the number of generations, while Google AI Plus/Pro/Ultra subscribers can get higher quotas.

Lyria 3 Project Address

  • Project Official Website：https://deepmind.google/models/lyria/

Lyria 3 Application Scenarios

  • Short Video Soundtracks: Quickly generate personalized background music for platforms such as Douyin, YouTube Shorts, and Instagram Reels to enhance content appeal.
  • Social Media Content: Automatically generate exclusive soundtracks for photos or videos such as travel vlogs, pet daily life, and food explorations, enhancing emotional expression.
  • Personal Entertainment Creation: Ordinary users can easily create personalized music works such as birthday blessing songs and anniversary theme songs, without any music background.
  • Podcasts and Audio Content: Generate opening and ending music or transition sound effects for podcast programs, audiobooks, audio ads, etc.
  • Games and Interactive Content: Generate customized background music and ambient sound effects for indie games, interactive stories, and virtual scenes.
  • Marketing and Brand Content: Enterprises can quickly generate original music that matches the tone for brand events, product launches, and advertising short films, reducing copyright costs.