Google Advances AI Media Models for Enterprise Solutions
Google has announced significant updates to its suite of AI models, aiming to solidify its position in the enterprise market. These enhancements span across various media formats, including **music**, **video**, **voice**, and **image** generation, all accessible through its Vertex AI cloud platform.
Lyria: Text-to-Music Model
Lyria, Google's text-to-music model, is now available in preview for select customers. This model offers an alternative to royalty-free music libraries, allowing users to create songs in diverse styles and genres, ranging from **jazzy piano solos** to **lo-fi tracks**.
Veo 2: Enhanced Video Creation
Veo 2, Google's video creation model, has received updates focusing on editing and visual effects customization. New features include the ability to remove background images, logos, and objects from existing videos. Furthermore, Veo 2 can extend video frames (e.g., converting landscape to portrait), adjust camera angles, and create timelapses and drone-style clips. It can also interpolate between specified beginning and end frames. These features are currently available in preview.
Chirp 3: Voice Cloning Technology
Chirp 3, Google's audio understanding model, now powers a voice-cloning feature called Instant Custom Voice. This feature, now generally available, can clone a voice using just 10 seconds of audio. Chirp 3 also underpins a new tool called Transcription with Diarization, available in preview, which separates and identifies speakers in recordings with multiple participants. Google emphasizes a "diligence" process to verify proper voice usage permissions for Instant Custom Voice, preventing abuse.
Imagen 3: Improved Image Generation
Imagen 3, Google's image generator, now delivers significantly better performance, particularly in removing objects and reconstructing missing or damaged portions of images. This enhancement boosts the model's utility in image editing and restoration tasks.
Safety Measures and Watermarking
All media generated by Imagen, Veo, and Lyria (excluding Chirp) are watermarked using Google's SynthID technology. Google emphasizes that all its generative AI models have built-in safeguards to protect against the creation of harmful content. The company also offers opt-out mechanisms for model training and an indemnity policy to protect Google Cloud and Vertex AI customers from AI-related copyright disputes.
These updates demonstrate Google's commitment to providing comprehensive and powerful AI tools for the enterprise market, directly competing with platforms like Amazon's Bedrock.
Source: TechCrunch