Google DeepMind and Darren Aronofsky debut AI-powered film ANCESTRA

Krea AI launches Krea 1 image model for high realism

SAG-AFTRA strike ends with new AI protections for performers

Kalshi airs AI-generated commercial during NBA Finals

The production developed new Veo capabilities, including motion matching for precise camera movements, "add object" functionality to seamlessly blend AI-generated babies into live-action footage, and enhanced personalization features.

Google's workflow combined multiple AI models—Gemini for prompt development, Imagen for concept art generation, and Veo for final video creation—then integrated these elements using traditional VFX compositing techniques.

The film required AI-generated baby imagery both in utero and during birth scenes, with Veo creating realistic newborn footage that traditional VFX would struggle to achieve without looking uncanny or requiring extensive time and resources.

McNitt's team used her father's personal photographs from the day she was born as reference material, feeding them into Gemini to create detailed aesthetic descriptions that became prompts for generating personalized imagery matching the film's intimate tone.

This marks the first of three experimental films in the DeepMind-Primordial Soup partnership, designed to test AI's role as a creative collaborator rather than replacement for human artistry.

The system uses Scenepacks of pre-built 3D environments that filmmakers can navigate and adjust using a simple game controller, eliminating the need for Unreal Engine expertise

Camera tracking options range from using just an iPhone to professional trackers

Their ARFX app works with displays of any size—from 4K TVs and projectors to professional LED walls—using the same software ecosystem throughout

The newly launched Infinite Studio AI tool can generate complete virtual environments in seconds, with depth mapping that allows camera movement within AI-created scenes

Unlike green screen workflows, ARwall focuses on in-camera visual effects (ICVFX), meaning what you capture is final pixel with no post-production compositing needed

Photorealism and Style Diversity: Krea 1 produces images with natural textures and avoids the soft, over-processed look common in other AI models. It supports a broad spectrum of styles, from photorealistic to abstract, retro to futuristic, and everything in between.

Speed and Resolution: Images are generated in about 8 seconds, with support for 1.5K native resolution and upscaling to 4K for professional use. Batch processing allows up to four images at once, and a real-time preview shows results as they’re created.

Style Reference and Custom Training: You can upload up to three reference images to guide the model’s style, mixing influences and adjusting their strength. Custom training on your own style or dataset is also supported, enabling highly personalized outputs.

Flexible Formats and Use Cases: Krea 1 supports a range of aspect ratios—landscape, square, and portrait—making it suitable for everything from product photography and concept art to editorial and marketing visuals.

Accessible Beta: The free beta lowers the barrier for experimentation, letting you try advanced features without complex sign-up or payment.

The tentative agreement includes critical AI protections that require explicit informed consent before any AI-driven replication of a performer's voice, face, or body, along with appropriate compensation for such usage.

The deal addresses three core performer demands: companies must get consent before creating AI versions of actors, compensate them fairly for AI usage, and clearly explain how AI doubles will be used.

Popular games like Destiny 2 and World of Warcraft left some NPCs unvoiced during otherwise voiced scenes due to the strike, while Call of Duty: Black Ops 6 characters were recast entirely.

The agreement now awaits final approval from SAG-AFTRA's National Board and subsequent membership ratification, with full contract details to be released upon completion.

Rapid AI workflow enabled PJ Ace to generate 15 usable clips from 300-400 iterations using detailed prompts that described each scene as if Veo 3 had no context of surrounding shots.

Google's Veo 3 delivered high-resolution output with advanced prompt engineering capabilities, allowing complex scene construction from simple text input without traditional cameras, actors, or physical sets.

Major networks approving this AI-native content for prime-time broadcast marks a significant shift in advertising standards and creative acceptance.

PJ claims the production cost represents a 95% reduction compared to traditional TV ads.