Particle.news

Download on the App Store

YouTube Brings DeepMind’s Veo 3 Fast to Shorts With Sound, Unveils New AI Creation Tools

Access starts in five countries, with tests of motion transfer, automated editing and speech-to-song plus SynthID labels and expanded likeness detection for provenance.

Overview

  • Veo 3 Fast is rolling out in the U.S., U.K., Canada, Australia and New Zealand, generating 480p clips with lower latency and synchronized audio from text prompts for free in Shorts.
  • New Veo-powered options coming to Shorts include applying motion from one video to a photo, stylizing footage with looks like pop art or origami, and inserting objects via text descriptions.
  • YouTube’s Speech to Song remix uses DeepMind’s Lyria 2 to turn dialogue from eligible videos into soundtrack songs, with testing beginning soon and wider U.S. access in the coming weeks and attribution to the original creator.
  • 'Edit with AI' assembles a first draft from raw footage by selecting highlights, adding music and transitions, and offering reactive voiceover in English or Hindi, now in experiments on Shorts and the YouTube Create app.
  • YouTube will apply SynthID watermarks and content labels to AI-generated media and is opening its likeness-detection tool to YouTube Partner Program creators to help find and manage unauthorized AI uses of their face; the company also touted $100 billion in creator payouts since 2021.