OpenAI abruptly discontinued its Sora video generation app and API on March 24, citing unsustainable compute costs and a strategic pivot to robotics. The $1B Disney partnership collapsed — no money changed hands. Sam Altman told staff the company is winding down ALL video products. ChatGPT will no longer support video functions. This is the single biggest AI video story of 2026. It validates Chinese competitors (Kling, Seedance, Hailuo) as the new leaders. ByteDance rolled out Seedance 2.0 globally via CapCut the SAME WEEK. The power has permanently shifted.

Three deep dives dropped this week that Founding Members got first: 🔧 Building a Local 4K AI Video Pipeline — Full NVIDIA + ComfyUI technical breakdown. No cloud costs, no waiting in queues. Your GPU, your footage, your workflow. 🎭 Real-Time AI Avatars with Runway Characters — How world models meet interactive media. Step-by-step build from zero to working avatar. 🎬 Daniel Kwan's AI Roadmap from SXSW — The "Everything Everywhere All at Once" director spent 3 years going deep on AI. His framework for filmmakers is the most practical thing I've seen from someone actually making films. We're past the halfway mark to 50 Founding Members. Once we hit 50, the price goes up and the door closes at this rate. So join TODAY! $29/month, locked for life → https://www.skool.com/aiography/classroom

Stop Fighting Your AI Video Tool — Use Structured Controls Instead If you've ever spent 20 minutes rewriting a prompt to get a dolly-in shot, this one's for you. Kling 3.0 dropped this week with an AI Director mode — and it's a different philosophy than what you're used to. Instead of describing camera moves in text and hoping the model interprets it correctly, you specify the shot: • Camera movement: Dolly, pan, tilt, crane (pick from a menu, not a paragraph) • Shot type: CU, MCU, wide, OTS (cinematographer language, not prompt engineering) • Scene transitions: Cut, dissolve, match cut (built-in storyboarding) • Character consistency: Plan a multi-shot sequence, lock the character across clips This isn't better or worse than Sora/Runway/Veo — it's built for a different workflow. If you're pre-visualizing a scene or prototyping a sequence, structured controls are faster. If you're exploring or generating B-roll, natural language prompts are more flexible. The lesson: stop forcing one tool to do everything. Match the tool to the task. Discussion question: Are you a "structured controls" filmmaker or a "natural language" filmmaker — and does your current tool match that? Founding Members get the full tool comparison breakdown (Kling vs Sora vs Runway vs Veo) with workflow decision trees and when to use which approach — skool.com/aiography/classroom

State Media Just Deployed AI Animation as Geopolitical Messaging. Chinese state media CCTV released "The Eagle vs. The Persian Cat" this week — an AI-animated political allegory using animal characters to comment on Gulf tensions. Eagle = USA, Persian cat = Iran. Full AI animation pipeline. Set aside the politics for a second. Watch the craft: - Character design that carries emotional weight — anthropomorphized animals with clear personality - Narrative economy — tells a complex geopolitical story in under 3 minutes - Production polish — lighting, composition, camera movement that rival traditional animation - Emotional manipulation — you know exactly who you're supposed to root for (regardless of whether you agree) This is what happens when AI video generation stops being a creator toy and becomes a communication tool for institutions. Government propaganda offices are now using the same AI animation pipelines indie filmmakers are learning on YouTube tutorials. That's both impressive and unsettling — which is exactly why it's this week's Video of the Week in the newsletter. Questions for the community: 1. When state media can produce AI animation this polished, what does that mean for independent filmmakers competing for attention? 2. Does the source of an AI film (state media vs. indie creator) change how you evaluate the craft? 3. If you were teaching AI filmmaking ethics, would you show this as a case study? Why or why not? Watch it here: The Eagle vs. The Persian Cat on YouTube Founding Members: Head to the FM section for this week's deep-dive workshop on Adobe's new multi-model video editor — five AI video engines (Pika, Sora, Luma, Veo, Firefly) in one workspace. Link: skool.com/aiography/classroom

Sunday night at the Oscars, Conan O'Brien called himself "the last human host." Will Arnett got a standing ovation for: "Animation is more than a prompt. It's an art form and it needs to be protected." Four days earlier, Netflix paid $600M for Ben Affleck's AI post-production startup. So what's the real message? The tension I'm seeing: • Public stance: AI threatens human craft, must resist • Private reality: Massive investments in AI infrastructure • The gap between the stage speech and the boardroom deal • Editors, VFX artists, animators watching from the middle Here's my read: Hollywood wants to be seen resisting AI while adopting it behind closed doors. The Oscars are the public face. The $600M deals are the private truth. For those of us building with AI tools, this matters. You're on the right side of where the industry is actually going — but don't expect applause from the stage. Question for you: Do you think Hollywood's public AI resistance helps or hurts independent creators who are already using these tools? Does the anti-AI rhetoric protect jobs or just delay the inevitable conversation? I've lived through every tech disruption in this industry — analog to digital, linear to nonlinear, broadcast to streaming. The pattern is always the same: public fear, private adoption, then the tools become infrastructure. What's your take? Are we watching resistance or theater?