Up.D-AI-TE / no.3

Subjective review of the news from the AI world that turns us on

Welcome to another UP.D-AI-TE! The AI landscape continues to evolve at breakneck speed, with groundbreaking developments reshaping how we create, produce, and think about audiovisual content. This month brings revolutionary breakthroughs from tech giants, legal battles defining AI’s boundaries, and tools that blur the line between human creativity and artificial intelligence. Let’s dive into what’s transforming our industry!

1. Google DeepMind’s “ANCESTRA”: Veo AI Meets Hollywood in Groundbreaking Film Partnership

Google DeepMind has partnered with acclaimed director Darren Aronofsky and filmmaker Eliza McNitt to create “ANCESTRA,” the first professional film to seamlessly blend live-action cinematography with Veo AI-generated footage. Premiering at Tribeca Festival 2025, this collaboration represents a watershed moment for AI in professional filmmaking.

Revolutionary production workflow combining AI and traditional filmmaking:

  • Gemini AI for prompt development using real family photos as reference material
  • Imagen for key concept art defining overall look, style, and mood
  • Veo for animating generated images and creating final shots with precise motion control
  • Traditional VFX pipeline for seamless compositing and color grading

Breakthrough capabilities developed for professional production:

  • Personalized video generation – Fine-tuned Imagen models to match specific artistic direction across scenes
  • Motion-matched video generation – Veo tracks virtual camera movements for precise directorial control
  • “Add object” capability – Inserting AI-generated elements into live-action footage while maintaining consistency
  • Blending traditional filmmaking and generative video – Composing actor performances with generated realistic imagery

This collaboration is the first of three films being made with Primordial Soup, Aronofsky’s new venture dedicated to storytelling innovation. The partnership involved Google DeepMind assembling a multidisciplinary creative team of animators, art directors, designers, writers, technologists and researchers working with more than 200 experts in traditional filmmaking and production.

Eliza McNitt described her experience: “Veo is another lens through which I get to imagine the universe around me.” The film portrays a mother’s love as a cosmic, life-saving force, inspired by the dramatic events of McNitt’s own birth.

The collaboration developed new Veo capabilities including personalized video generation, precise motion matching, and the ability to blend live-action and generative footage, demonstrating how generative AI can complement existing creative and production workflows.

Source: ANCESTRA: Behind the Scenes – Google DeepMind

2. Apple Intelligence Foundation Models: The Privacy-First AI Revolution for Content Creators

Apple has unveiled its Foundation Models at WWDC 2025, introducing a revolutionary approach to AI that could transform content creation workflows. The system features a 3-billion-parameter on-device model optimized for Apple Silicon, alongside a server-based mixture-of-experts model designed for Private Cloud Compute.

Game-changing features for content creators:

  • Complete privacy protection – No user data leaves the device, perfect for sensitive creative projects
  • 15 language support – Global content creation without cloud dependency
  • Vision + language capabilities – Process images and text simultaneously for richer creative applications
  • New Foundation Models Framework – Developers can build AI features directly into apps using Swift

Why this revolutionizes audiovisual production: Apple’s approach of 2-bit weight compression delivers lightning-fast inference without quality loss, making this ideal for on-set creative tools. The “guided generation” feature ensures reliable, structured outputs perfect for production workflows. The system can generate 5-second videos at 1080p resolution while maintaining complete privacy protection.

Apple’s Foundation Models framework gives developers access to the on-device language model through an intuitive Swift approach called guided generation, enabling app developers to create production-quality generative AI features with rich data structures and strong type guarantees.

Source: Apple Intelligence Foundation Models – Technical Report

3. TikTok Symphony Expands: Next-Gen AI Tools for Brand Content Creation

TikTok unveiled major updates to its Symphony generative AI suite at Cannes Lions 2025, introducing three powerful new tools designed specifically for brand content creation. The platform is positioning itself as the creative partner for the entire marketing ecosystem, from global brands to small businesses.

New Symphony capabilities transforming content creation:

  • Image to Video – Transform static product photos and brand assets into scroll-stopping 5-second TikTok clips instantly
  • Text to Video – Generate engaging video concepts from simple text prompts, perfect for rapid ideation and testing
  • Showcase Products – Digital avatars model products, clothing, or apps, bringing static images to life with human interaction

TikTok is bringing Symphony directly into Adobe Express and WPP Open, meeting marketers where they already work. This represents a shift from platform-specific tools to integrated creative workflows.

Andy Yang, Global Head of Creative and Brand Products, stated: “We’re entering a new era of creativity, one where ideas move at the speed of culture, and where AI doesn’t replace imagination but rather accelerates it.” All Symphony content is automatically labeled as AI-generated and undergoes multiple safety reviews.

Aubrey Cattell, VP Developer Platform & Partner Ecosystem at Adobe, noted: “Introducing Symphony capabilities to Adobe Express users unleashes their potential, simplifying TikTok-first content generation while broadening their creative possibilities.”

Rob Reilly, Chief Creative Officer at WPP, added: “With TikTok’s Symphony Suite, we’re giving our creatives even more firepower to push boundaries and experiment for our clients.”

Source: TikTok Symphony Updates – Cannes Lions 2025

4. Seedance 1.0: ByteDance’s Ultra-Fast Video Generation Breakthrough

ByteDance has unveiled Seedance 1.0, a revolutionary video generation model that achieves unprecedented speed and quality in AI video creation. The model generates 5-second 1080p videos in just 41.4 seconds, representing a 10× inference speedup through advanced distillation strategies and system-level optimizations.

Technical breakthroughs reshaping video production:

  • Multi-source data curation with precision video captioning for comprehensive learning across diverse scenarios
  • Native multi-shot generation capacity enabling coherent narrative storytelling with consistent subject representation
  • Multi-dimensional reward mechanisms through video-specific RLHF for superior performance
  • Interleaved multimodal positional encoding for enhanced prompt following and motion plausibility

Advanced capabilities:

  • Superior spatiotemporal fluidity with structural stability
  • Precise instruction adherence in complex multi-subject contexts
  • Ultra-fast inference making real-time creative iteration possible
  • Native multi-shot narrative coherence for consistent storytelling

Seedance 1.0 integrates several core technical improvements including an efficient pre-training paradigm that enables multiple features such as interleaved multimodal positional encoding, native multi-shot generation capacity, and multi-task modeling. The model achieves considerable performance improvements through carefully-designed post-training optimization leveraging fine-grained supervised fine-tuning and video-specific RLHF.

Source: Seedance 1.0: Video Generation Model – ByteDance

5. Disney and Universal Sue Midjourney: Hollywood’s Copyright Battle Over AI Image Generation

Disney and Universal have filed a federal lawsuit against AI image generator Midjourney, alleging the company operates as a “bottomless pit of plagiarism” by creating unauthorized copies of characters including Darth Vader, Frozen’s Elsa, and the Minions. The lawsuit was filed in federal district court in Los Angeles.

Key allegations in the copyright case:

  • “Innumerable” copies of Disney characters including Star Wars’ Yoda and Marvel’s Spider-Man, Incredible Hulk, and Iron Man
  • $300 million revenue generated by Midjourney last year alone
  • “Soon-to-be-released video service” planned by the AI company
  • Direct copying without transformation – characters appear in new locations or backgrounds without creative changes

Legal experts are divided on the case’s strength – while Syracuse University’s Shubha Ghosh argues Midjourney creates direct copies of copyrighted characters without creative transformation, Hall Estill’s Randy McCarthy warns the litigation faces complex issues around terms of service and fair use that courts must resolve.

Disney maintains that AI-enabled piracy remains piracy regardless of technology, though the company supports responsible AI creativity tools.The lawsuit reflects Hollywood’s complex relationship with AI, coming two years after actors and writers struck demanding AI protections, while current films like Emilia Pérez and The Brutalist use AI for voice alteration, and the technology is employed to de-age actors like Tom Hanks and Harrison Ford.

Source: Disney and Universal sue AI firm Midjourney – BBC News

6. Luma AI’s Modify Video: Revolutionary Post-Production Tool Transforms Existing Footage

Luma AI has launched Modify Video, a groundbreaking AI tool that allows creators to completely transform existing video footage while preserving the original performance, motion, and camera work. This technology eliminates the need to “start from scratch” when changing environments, lighting, or visual elements in post-production.

Game-changing capabilities for video editing:

  • Motion & Performance Capture + Puppeteering – Extract full-body, facial, or lip-sync motion to drive new characters or environments
  • Restyle, Retexture, and World Swapping – Transform scenes (garage → spaceship, day → night) while keeping motion intact
  • Isolated Motion VFX & Scene Element Edits – Edit individual elements like wardrobe, faces, or props without affecting the entire shot
  • Three structured presets: Adhere (retexturing), Flex (balanced transformation), Reimagine (full creative freedom)

Revolutionary workflow advantages:

  • Preserves motion and action using advanced performance signals like pose, lip sync, and facial tracking
  • Multiple output variants from the same base motion for rapid client feedback
  • Prompt-optional interface – control outputs with visuals, not guesswork
  • Native resolution support including 16:9 (720p) for seamless integration

External blind evaluations conducted with external raters showed that Luma Modify Video consistently outperformed Runway V2V across key criteria. Luma excelled in retaining motion, facial animation, and temporal consistency, providing a smoother, more coherent viewing experience.

Luma AI believes professional creatives should be able to reimagine environments, lighting, and texture without losing the integrity of the performance, motion, camera or character. With Modify Video, creators can keep what matters and evolve everything else.

Source: Luma AI Modify Video – Revolutionary Post-Production Tool