Disney bets $1B on AI and Sora

Disney goes all in on AI with Sora

McDonald's AI Christmas fail

Beeble ships local GPU relighting app

sync.'s new react-1 model edits character performances

200+ licensed characters - Sora will generate short, user-prompted social videos using more than 200 animated, masked, and creature characters from Disney, Marvel, Pixar, and Star Wars, including costumes, props, vehicles, and iconic environments. ChatGPT Images will also generate images from the same IP library. The agreement explicitly excludes any talent likenesses or voices.

Disney+ integration - Curated selections of Sora-generated fan videos will stream on Disney+, and OpenAI will collaborate with Disney to build new Disney+ subscriber experiences using OpenAI's models. Sora and ChatGPT Images are expected to start generating Disney character content in early 2026.

Enterprise deployment - Disney will become a major OpenAI customer, using OpenAI's APIs to build new products and tools for Disney+ and deploying ChatGPT internally for employees. The financial structure includes a $1 billion equity investment plus warrants, subject to definitive agreements and corporate approvals.

Safety and rights framework - Both companies committed to maintaining controls to prevent illegal or harmful content generation, respect content owner rights in model outputs, and protect individual voice and likeness rights. OpenAI pledged to implement age-appropriate policies and safety controls across the service.

Included characters - The licensed library spans Mickey Mouse, Minnie Mouse, Lilo, Stitch, Ariel, Belle, Cinderella, Simba, Mufasa, plus characters from Encanto, Frozen, Inside Out, Moana, Monsters Inc., Toy Story, Up, Zootopia, and animated versions of Marvel and Star Wars characters including Black Panther, Captain America, Deadpool, Groot, Iron Man, Darth Vader, Luke Skywalker, Leia, the Mandalorian, and Yoda.

Agentic story development - Feed Eddie a URL in Rough Cut mode, and it automatically pulls key messages, brand positioning, and background context into the edit—so early cuts are aligned with client narratives from the start. (Treatment input coming soon!)

Extended rough cuts - Rough cut limit now stretches to 40 minutes, giving YouTubers, documentarians, and webinar creators room to edit long-form projects in one pass without external workarounds.

Smarter B-roll logging - Eddie now analyzes both visuals and dialogue in B-roll, making background footage searchable by what’s seen and said—surfacing contextual moments that often get missed.

On-prem processing - The app runs entirely on local GPUs, keeping footage in-house for studios managing sensitive projects. No uploads, no cloud bottlenecks, no credit limits.

4K relighting with full PBR outputs - Renders sequences up to 1 hour (100,000 frames) at 4K resolution. Exports multi-channel 16-bit EXRs with complete PBR AOV passes: normals, base color, roughness, metallic, specular, alpha, and depth in a single file. Imports directly into Nuke, Blender, and Unreal Engine.

SwitchLight 3.0 model - The new version processes multiple frames simultaneously for flicker-free, temporally consistent results. Trained on a dataset ten times larger than before, with sharper facial definition, better surface textures, and improved background stability.

Built-in tools - Includes a render queue for batch jobs, deflicker controls, and the Beeble Editor, a 3D editor for real-time relighting with physically accurate HDRI and point lights.

Pricing - Indie plan runs $500/year or $60/month. Standard plan is $3,000/year or $400/month for full commercial use.

Beyond lip sync - react-1 "learns from your uploaded audio and reanimates the entire face," editing facial expressions, head movements, and timing across the performance, not just lip sync. You upload existing footage, provide new audio, guide the emotional direction, and the model generates a new performance while maintaining the actor's identity and style.

Emotion-level control - The interface offers selectable emotional reads like surprised, angry, disgusted, sad, happy, and neutral, letting editors "explore different reads with the click of a button" and change the emotional beat of a performance in post.

Localization angle - sync. positions this for dubbing workflows, claiming react-1 "doesn't just localize the lines it localizes the entire performance," targeting global content pipelines where emotional performance needs to match new language or script changes.

Post-production integration - Built on sync.'s existing lip sync API infrastructure, react-1 works "on any video content in the wild—across movies, podcasts, games, and even animations" and is available via API and self-serve web product with a "Start for free" onboarding flow.