I tested Sora's new 'Character Cameo' feature, and it was borderline disturbing (www.zdnet.com)

OpenAI’s Sora app has launched a “Character Cameo” feature that turns brief video clips of animals, toys, statues or original creations into reusable, named AI characters with autogenerated usernames, personality descriptions and optional behavioral “restrictions.” The app animates and voices those characters in new short videos, lets creators edit the character metadata, and controls who can reuse them (just you, approved users, mutuals, or everyone). Sora remains invite-only generally but is temporarily open without an invite in the US, Canada, Japan and Korea; outputs are watermarked and the feature explicitly blocks generating cameos of real humans. The addition is significant because it lowers the bar to producing shareable, character-driven synthetic media—fuel for social feeds and creative storytelling—while also surfacing technical and safety questions. Sora can generalize from a single clip to new movements and speech (the author saw a raccoon “Rooftop Rascal” stash trinkets and do a clumsy dougie, and a “Lava Lizard Sage” statue animate and speak in a booming, uncanny voice). That capability demonstrates powerful motion/voice synthesis and identity-tagging, but also amplifies risks: anthropomorphism and uncanny outputs, potential for disturbing content, reuse by others, moderation and consent challenges. The feature’s metadata, restriction toggles and watermark help, but creators and platforms will need policies to manage misuse and community impact as these synthetic characters proliferate.