Google introduces a new text-to-music generator Lyria, which can be used via Vertex AI to add music to videos created with Veo 2.
This week, Google Cloud Next is taking place, the high mass of Google’s cloud division, and thus it’s raining AI-related announcements. These don’t revolve solely around Gemini. AI generation is permeating various forms of content creation. After generating text, images, and video, you can now also create music with the help of AI thanks to Google’s latest model Lyria.
Creating Music with Lyria
Lyria will initially be available through Vertex AI, the platform where all of Google’s AI models come together, it announces in a blog. With Lyria, users can generate music based on text instructions. This creates the possibility to make a complete multimedia production from a single text prompt: from image and video to speech and music.
Lyria is currently available in preview and you need to manually activate the model. The model can create audio in different styles and is positioned as a tool for, among others, marketing campaigns, podcast production, and digital content creation. Google shares a YouTube link with a sample created by Lyria.
Image, Video, and Speech
The introduction of Lyria is new, but Google is also further refining the existing models for media creation. Imagen 3 and Veo 2 are getting more extensive editing options, including the ability to remove or add objects to the image. The introduction of new camera angles makes it possible to introduce visual effects, and with interpolation, you can refine transitions to make your videos look professional.
With Lyria, you can add music to your video, and Chirp provides speech. Chirp 3 introduces Instant Custom Voice to create a custom voice based on just 10 seconds of audio material. Additionally, there’s a transcription update that automatically distinguishes speakers in recordings. Both functions are available in preview via Vertex AI.
Google emphasizes that responsible use and privacy are central. It uses digital watermarks to label AI-generated content, and built-in safety filters prevent the generation of inappropriate or harmful content. Furthermore, Google promises that customer data is not used to train models.