Today in Generative Media

'AI slop' summer; Lionsgate's AI will edit for PG-13; Epic's AI to write game dialogue

News and Opinion

• Enjoy ‘AI slop’ summer. What’s coming next is worse (Fast Company)

• Lionsgate Boss Says Studio Can Use AI to Adjust a Movie’s Rating and Create Kid-Friendly Cuts: ‘Do It Anime, Make It PG-13. Three Hours Later, I’ll Have the Movie’ (Variety)

• AI Prompts Will Soon Let a 10-Person Team Build a Game Like Breath of the Wild Where the AI Is Doing All the Dialogue and You Just Write Character Synopsis, Tim Sweeney Predicts (IGN)

• Runway Partners with AMC Networks Across Marketing and TV Development (RunwayML)

• FDA rushed out agency-wide AI tool—it’s not going well (Ars Technica)

• Anthropic co-founder on cutting access to Windsurf: ‘It would be odd for us to sell Claude to OpenAI’ (TechCrunch)

• AI reveals hidden language patterns and likely authorship in the Bible (Phys.org)

• My AI Skeptic Friends Are All Nuts (Thomas Ptacek’s blog)

Software

• The Future of Voice: Bland’s New Breakthrough TTS Engine (bland.ai)

• We have heard your asks, and worked with Kling team to deliver sub-minute video generation on it's latest and most affordable model, Kling 2.1 Standard (X) [fal.ai]

• UniWorld-V1, a unified framework for understanding, generation, and editing. (GitHub)

• Portraits: Expert knowledge, delivered by AI coaches. (Google Labs)

• We present SceneFactor, a diffusion-based approach for large-scale 3D scene generation that enables controllable generation and effortless editing. (GitHub)

• Our latest Gemini 2.5 Pro update is now in preview. It’s better at coding, reasoning, science + math, shows improved performance across key benchmarks (AIDER Polyglot, GPQA, HLE to name a few), and leads LM Arena with a 24pt Elo score jump since the previous version. (X)

Research

• Facial Appearance Capture at Home with Patch-Level Reflectance Prior (project page)

• Low-Rank Head Avatar Personalization with Registers (project page)

• FlowMo: Variance-Based Flow Guidance for Coherent Motion in Video Generation (project page)

• DCM: Dual-Expert Consistency Model for Efficient and High-Quality Video Generation (project page)

• Video World Models with Long-term Spatial Memory (project page)

• Native-Resolution Image Synthesis (project page)

• Image Editing As Programs with Diffusion Models (project page)

• ShotAdapter: Text-to-Multi-Shot Video Generation with Diffusion Models (project page)

• Temporal In-Context Fine-Tuning for Versatile Control of Video Diffusion Models (project page)

• Reconstructing scenes from video captured on any egocentric device (project page)

• PBR-SR: Mesh PBR Texture Super Resolution from 2D Image Priors (project page)

• Geometry Distributions (project page)

• High-Fidelity Lightweight Mesh Reconstruction from Point Clouds (CVPR Open Access)

• Solving Inverse Problems with FLAIR (arXiv)

• How much do language models memorize? (arXiv)

Misc

• 3DV 2026: Call for Papers