Today in Generative Media

Writers making AI sound human; UMG's AI voice cloning deal; Suno pays AI creators

News

• AI took their jobs. Now they get paid to make it sound human (BBC)

• Universal Music artists get access to AI voice cloning tool via UMGs new deal with tech startup SoundLabs (Music Business Worldwide)

• After raising $125M, AI music generator Suno is now paying its most popular creators (Music Business Worldwide)

• Databricks and Shutterstock are trying to remove the copyright risk from AI image generation (Fast Company)

• AI could help spread false and misleading information on Holocaust, UNESCO report warns (AP News)

• DC Comics cover artist accused of using AI and swiping character poses (BoingBoing)

• Runway Gen-3 Can Make AI Videos of ‘Photorealistic Humans’ (PetaPixel)

• 6 AI-powered Canva features every creative pro needs to try (Pocket-Lint)

• ElevenLabs unveils open-source creator tool for adding sound effects to videos (VentureBeat)

• Neurapix Launches the World’s First Personalized AI Cropping Tool (PetaPixel)

• Temporary Stable Diffusion 3 Ban (CivitAI)

Software

• Introducing the research preview of our foundation model, Character-1. Available today at hedra.com (on desktop and mobile). * Infinite duration (30s for open preview) * 90s generated per 60s (if our H100 supply holds) * Expressive talking, singing, rapping characters (X)

• Today is the start of a new chapter! Super excited to announce the start of comfy.org w/ a super star lineup. It includes comfyanonymous(ComfyUI), mcmonkey4eva(SwarmUI), Dr lt Data(UI Manager), and a few star developers. Our goal is to take ComfyUI to the next level (X)

• Open Sora has taken a step ahead of OpenAI, delivering another stunning upgrade for the open-source community: complete open-source weight codes! It can generate up to 16 seconds of 720p HD visuals with incredibly realistic character rendering and cinematic zoom effects. (X)

Research

• L4GM: Large 4D Gaussian Reconstruction Model (project page)

• 📏Metric3D v2: A Versatile Monocular Geometric Foundation Model for Zero-shot Metric Depth and Surface Normal Estimation (project page)

• RetinaGS: Scalable Training for Dense Scene Rendering with Billion-Scale 3D Gaussians (arXiv)

Misc

• DepthAnythingV2 Small F16 perf on iPhone 12 an old ass phone, while recording (X)

• ChatGPT crossed 1M users in 5 days. Luma crossed it in 4 days. (X)

• Sure starting to feel like this bubble is getting ready to pop. Nvidia is the most valuable company in the world—worth more than Microsoft, more than Apple, more than *every stock in the UK combined*. Yet McDonald's and others are canceling AI programs because they don't work. (X)

• Something I think the AI dorks will never understand is how half the reason most real filmmakers want to make films is because that’s a FUN thing to do. Working with other artists and actually MAKING something is a joyful act and without the process the result is hollow. (X)

• "Write what you can produce" has traditionally guided filmmakers to write small and contained, grounded. Get your friends to act, shoot in your basement. Improvise like crazy. It'll be interesting to see what happens when these restrictions loosen up. When your amateur movie can have an UFO chase and robots destroying your hometown. (X)