xAI's grok-imagine-video-1.5-preview is now live via the xAI API and climbing toward broader consumer rollout, and it just grabbed the #1 spot on the Image-to-Video Arena leaderboard. It animates a single still with camera moves, physics, and built-in synchronized audio at up to 720p, and you can chain clips for longer consistent sequences. Pricing runs $0.08/sec at 480p and $0.14/sec at 720p — cheap enough to actually test in a real pipeline.
Lightricks' open-source LTX-2, built with NVIDIA, is now natively supported in ComfyUI and generates up to 20 seconds of 4K video at 50fps with synced dialogue, Foley, and music in a single pass — on one consumer GPU. NVIDIA's NVFP4 path runs it up to 3x faster with 60% less VRAM. If you're self-hosting, this is the open model that finally gets sound right without a separate audio stage.
YouTube's January enforcement wave was its largest ever against AI slop — 16 channels terminated, 4.7 billion lifetime views and 35 million subscribers wiped, roughly $10M in annual ad revenue erased. The line isn't AI versus no-AI; it's whether a human made real editorial decisions or whether it's identical templated, AI-voiced output at volume. Anyone running a faceless channel needs commentary, narrative, or genuine editorial steering to stay monetized.
Stable Video Infinity 2.0 Pro now supports the Wan 2.2 base model, letting you start from a single image and generate theoretically unlimited-length clips with natural scene transitions. New custom nodes pair SVI's motion continuity with Wan 2.2 First/Last Frame control, and a recent update lets you load an existing video and continue from it via a last frame, a prompt, or both. The shipped workflows handle up to 7 clips, but that's not a hard ceiling.
Alibaba's 15-billion-parameter HappyHorse-1.0 climbed anonymously to #1 on Artificial Analysis before being unmasked, hitting 1389 Elo in text-to-video and beating Seedance 2.0 by nearly 115 points. It's a unified Transformer generating synced video, dialogue, ambient sound, and Foley in one forward pass, with native lip-sync across seven languages. It's already live on fal, so you can put it head-to-head against your current stack today.
ComfyUI shipped App Mode, App Builder, and ComfyHub — a one-click system that swaps the node graph for a clean inputs-and-outputs UI you can hand to clients or collaborators who've never touched a node. App Builder lets the author pick which inputs and outputs surface, and ComfyHub is the public showcase for sharing finished apps. For creators selling services, this is the productization layer ComfyUI was missing.
Platform AI-disclosure rules have hardened: TikTok now scans uploads for C2PA metadata and auto-applies its "AI-generated" label, Meta unified Instagram and Facebook AI policies in February requiring in-post disclosure separate from the caption, and YouTube keeps AI content fully monetizable with proper disclosure. The practical headache — you may need to label a video on TikTok that needs no label on Instagram or YouTube. Build disclosure into your publishing checklist per platform.
ByteDance suspended the global rollout of Seedance 2.0's API back in mid-March after Disney, Warner Bros. Discovery, Paramount, and Netflix sent cease-and-desist letters alleging the model was effectively pre-loaded with copyrighted character designs. There's no clear timeline for the overseas API returning, and a first-half-2026 recovery looks unlikely. If your roadmap was banking on Seedance access, route around it — for now it's a benchmark headline you can't actually build on.
