The end of 2024 in Generative Media!

AI isn't genius, we are; 2025 full of AI, copyright issues for big tech; Meta's AI bots

News and Opinion

• A.I. Isn’t Genius. We Are. (New York Times)

• Good news! 2025 will be full of AI! (Washington Post)

• Tech companies face tough AI copyright questions in 2025 (Reuters)

• Facebook planning to flood platform with AI-powered users (Futurism)

• The Critical Necessity of Immediate Regulation and Labeling of AI-Generated Media (Fstoppers)

• How Generative AI And LLMs Are Reinventing Our Vocabulary Such That We Might Lose Our Grasp On Human Languages (Forbes)

• The 19th-Century Technology That Threatens A.I. (New York Times)

• DeepSeek’s new AI model appears to be one of the best ‘open’ challengers yet (TechCrunch)

• AI data centers are becoming 'mind-blowingly large' (ZDNet)

• AI upscaling killed native graphics gaming. We’re better off for it (PCWorld)

• 2024 Lookback: A watershed year for generative AI in Hollywood (Variety, paywalled)

• 20 Media Executives Offer Their Predictions for 2025 (AdWeek)

• 10 Predictions for AI, Media & Entertainment in 2025 (Peter Csathy)

• Where will AI be at the end of 2027? A bet (Gary Marcus and Miles Brundage)

Software

• ChatRTX is a demo app that lets you personalize a GPT large language model (LLM) connected to your own content—docs, notes, images, or other data... And because it all runs locally on your Windows RTX PC or workstation, you’ll get fast and secure results. (NVIDIA)

• This is the official implementation of our NeurIPS 2024 paper "HDR-GS: Efficient High Dynamic Range Novel View Synthesis at 1000x Speed via Gaussian Splatting" (GitHub)

• fal-ai/video-upscaler (fal.ai) The video upscaler endpoint uses RealESRGAN on each frame of the input video to upscale the video to a higher resolution.

• We've just released the training code for OminiControl! (X)

• ComfyUI just added Hunyuan video model support for 8gb VRAM cards! (X)

Research

• DAS3R: Dynamics-Aware Gaussian Splatting for Static Scene Reconstruction (project page)

• Switti: Designing Scale-Wise Transformers for Text-to-Image Synthesis (project page)

• Generating Editable Head Avatars with 3D Gaussian GANs (project page)

• Dust to Tower: Coarse-to-Fine Photo-Realistic Scene Reconstruction from Sparse Uncalibrated Images (arXiv)

• Humans as a Calibration Pattern: Dynamic 3D Scene Reconstruction from Unsynchronized and Uncalibrated Videos (arXiv)

• ResGS: Residual Densification of 3D Gaussian for Efficient Detail Recovery (arXiv)

Misc

• Nike just dropped an ad made with a hybrid workflow that combines AI and traditional VFX. (X)

• Wait, so we now have a 100% open source model that's better than GPT-4o?! (X)

• DeepSeek v3 is an order of magnitude cheaper because it likely trained on frontier model outputs, in obvious violation of ToS ToS laundering by training on DeepSeek outputs is impossible to prevent. Does not bode well for economics of training frontier models (X)

• worth remembering what you're signing up for. Large generative models have the ability to deeply influence you. (X)