Black Forest Labs just launched the FLUX Pro Finetuning API, allowing creators to customize flagship FLUX Pro models with as few as 1 to 5 reference images. The custom finetunes reliably preserve unique characters or visual styles at up to 4 megapixels and work seamlessly across the full FLUX tool suite.
Luma AI just released Ray2, its next-gen video model trained with 10x the compute power of Ray1. It brings coherent physics, cinematic camera controls, 1080p output with 4K upscaling, and seamless video extensions up to 30 seconds. It is live now for paid Dream Machine subscribers on web and iOS.
Freepik just launched AI video generation with sound directly inside its AI Suite, supporting multiple modes and formats. Instead of generating silent clips and hunting for soundtracks in separate apps, creators can now get complete audiovisual clips in a single workflow.
Stability AI and NVIDIA just rolled out SPAR3D at CES, turning single 2D images into detailed 3D objects in just 0.7 seconds. The architecture pairs a point diffusion model with a triplane transformer, letting creators edit point clouds and rebuild meshes in 0.3 seconds. Weights are open on Hugging Face for RTX AI PCs.
Polish director Patryk Vega just set his $15M biopic Putin for a Jan 10 premiere across 64 countries. The production used AI visual effects to superimpose the Russian president's real face onto an actor over his body language. Why it matters: It brings deepfake-level AI straight into feature-length cinema, sparking fierce ethical debate along the way.
Midjourney just rolled out Pinterest-inspired moodboards and multi-profile personalization, letting creators train custom aesthetic models directly from curated image collections. Users now only need 40 image ratings to build a profile instead of 200, and can seamlessly switch between distinct named styles for different creative projects.
Google just unveiled Veo 2, a new video model that generates clips up to 4K resolution spanning several minutes while understanding cinematic prompts like lens focal lengths and depth of field. Alongside it, an upgraded Imagen 3 launched globally in ImageFX, also powering a new image-remixing experiment called Whisk. Veo 2 is currently rolling out via a waitlist on VideoFX.
NYU Tisch just partnered with Runway to embed its AI video tools into a new graduate virtual production course this spring. Based at the Martin Scorsese Virtual Production Center, film students will experiment with generative video across the entire workflow, from initial concepting to final thesis projects.
Pika just launched Pika 2.0, upgrading visual fidelity and text alignment across its video generator. The headline update is Scene Ingredients, a new feature allowing creators to upload reference images of people, places, and objects to maintain visual consistency across generated shots.
Manchester City and Puma just launched an AI kit generator powered by DeepObjects, inviting fans to design the club's 2026-2027 third kit. The Puma AI Creator gives users 15 credits to generate virtual jersey concepts and submit up to two entries by December 20th. The winning submission will mark the first AI-generated kit ever worn in an official match.
xAI just rolled out Aurora, a new autoregressive image generation model for Grok on X. Trained on billions of web examples, it tackles notoriously tough challenges like text rendering, logos, and photorealistic portraits, alongside native support for image editing inputs. It is now live in select countries and rolling out to all users within a week.
AMD just released and open-sourced Nitro Diffusion, a pair of one-step text-to-image models based on Stable Diffusion 2.1 and PixArt-Sigma. Trained on AMD Instinct MI250 accelerators, they match full-step visual quality in a single inference step for both data centers and edge PCs. Code and weights are now up on Hugging Face and GitHub.
OpenAI is gearing up to launch Sora v2 very soon, according to creative director Chad Nelson at a London keynote. The upgraded model can generate videos up to one minute long, adding image-plus-text and video-to-video capabilities alongside standard text prompts.
ElevenLabs just expanded its GenFM podcast creator to Studio, allowing anyone to turn articles, PDFs, and links into lifelike discussions between two AI co-hosts. Creators can edit the transcript, swap voices, and export the audio across 32 languages for commercial use. GenFM is also rolling out globally on the ElevenReader Android app.
AdCreative.ai just launched Product VideoShoot, an AI model that turns static product images into 16-second 1080p commercials. Powered by an ad-specific LLM trained on 840 million high-converting campaigns, it prioritizes performance metrics and brand guidelines over generic aesthetics. The feature is currently live in beta for Pro and Enterprise tiers.
Google just opened its flagship Imagen 3 generator to all Vertex AI customers, while launching video model Veo in private preview. Veo produces six-second clips from text or image prompts with localized editing support. Google is finally pushing its top creative models to enterprise clients, though Veo remains gated behind early access.
Hailuo AI just launched I2V-01-Live, a new image-to-video model built specifically to animate 2D illustrations. It delivers smoother character motion, subtle micro-expressions, and speaking animations across various art styles, turning flat concept art into lively video.
Fei-Fei Li’s startup World Labs just revealed its first project: an AI system that generates interactive, explorable 3D scenes from a single 2D photo. Users can navigate browser-rendered spaces with mouse and keyboard controls, complete with adjustable depth of field. It proves world models can maintain spatial consistency, though movement boundaries remain limited.
Adobe Research just unveiled MultiFoley, a video-aware audio generation method with multimodal controls. It lets creators transform sounds—such as making a typewriter sound like a piano or turning a cat's meow into a lion's roar—while automatically syncing effects to video timing. Why it matters: Foley sound design is notoriously tedious, and this could automate frame-accurate audio alignment.
A group called Sora PR Puppets just leaked access to OpenAI's Sora on Hugging Face, briefly letting the public generate 1080p video clips before getting shut down. The collective claims OpenAI exploited early testers for unpaid labor and PR hype rather than genuine artistic collaboration. Why it matters: Creator backlash and delays are making OpenAI's video ambitions significantly harder.
Stability AI just released three new ControlNet models for Stable Diffusion 3.5 Large: Blur, Canny, and Depth. These models allow creators to handle upscaling up to 16K, line art structure, and depth map guidance with high precision. They are free for non-commercial use and commercial use under $1M in annual revenue, supported by Comfy UI.
