Denmark wants consent for any realistic face or voice. Plus MIT-licensed open video with audio.
Jeff T. Thomas, who posts as Jeff Synthesized, beat roughly 2,500 entries with The Gifted, a three-minute AI proof-of-concept trailer, and the judges picked it unanimously. The prize is $100,000 cash plus $2.5 million in equity toward a feature that Range Media says will cost about $10 million and will not be AI generated. The trailer won the meeting, not the production.
Sony has asked platforms to pull more than 260,000 tracks imitating artists like Harry Styles and Adele, up from 135,000 in March. Spotify allows vocal impersonation only with the artist's permission, and its Artist Profile Protection lets artists approve releases before they reach their page. If your AI music or video borrows a famous voice, expect it to be found and removed.
Denmark's culture minister has presented a revised bill giving people a separate right over their voice and likeness, covering AI deepfakes and ordinary edits that could be confused with the real person. Parody, caricature and criticism are exempt unless the piece is harmful misinformation. The stated goal is fast platform takedowns, so get written consent before a recognisable person appears in client work.
The channel builds two AI influencers in Higgsfield AI Influencer 2.0: a rapper performed through motion reference, and a fictional actor from its own series who posts behind-the-scenes check-ins. The first gym check-in, made through Codex and the Higgsfield MCP, only felt natural once small pauses and glances were directed in. Higgsfield sponsors the video, and the creator says the money ideas are untested.
A viral TikTok has a parent shooting her kids against a plain wall and asking ChatGPT to make it a school picture, skipping $25 to $60 print packages. UK parents are also stripping photographers' logos off proofs with AI, and 58% of Association of Photographers members have lost commissioned work to AI. Your clients hold the same tools, so charge for judgement, not the render.
Kandinsky Lab open-sourced a 3B Lite and a 29B Pro model on 6 Oct, making five-second clips with synced 44 kHz audio and lip-sync. Code and weights are MIT, so client work is fine, unlike research-only open models. The repo's own timings: on an RTX 4090, Lite at SD takes about 7 minutes per clip and Pro at Full HD about 21.