The Render

OpenAI announced on Tuesday that it's discontinuing Sora — the most high-profile exit from the AI video race yet, and it landed right alongside its unwinding of the $1B Disney deal; meanwhile the open-source video model landscape just got a serious upgrade with Lightricks dropping LTX-2.3, and Runway's been quietly becoming a platform play — not just a video model — with a multi-model hub and world model ambitions that go well past the timeline. A lot moved this week.

What Happened: OpenAI on Tuesday said it will discontinue Sora, the generative AI video creation app it launched last year, without providing a reason for the decision. The Sora app shut down six months after going viral after its launch. The announcement comes just a day after OpenAI published a blog post about Sora safety standards. The Sora research team will continue to focus on world simulation research to advance robotics rather than consumer video. The Disney licensing deal — which covered more than 200 characters from Disney, Marvel, Pixar, and Star Wars — is also being wound down.

Why Is This Important? OpenAI entered the video space with enormous fanfare and is exiting after six months — that's a signal about compute economics, not creative appetite. For the production community, the practical fallout is limited: Sora was never close to a professional workflow tool. But the Disney deal implosion is telling. The IP rights conversation around AI video just got messier, not cleaner. And the vacuum OpenAI is leaving gets filled by Runway, Kling, and Veo — none of which have OpenAI's distribution. Watch who picks up those enterprise conversations.

What Happened: All of the world's best models are now available right inside Runway, including Kling 3.0, Kling 2.6 Pro, WAN2.2 Animate, GPT-Image-1.5, Sora 2 Pro, and more. Simultaneously, Runway launched GWM-1, an autoregressive model built on top of Gen-4.5 that generates frame by frame, runs in real time, and can be controlled interactively with actions — camera pose, robot commands, audio — coming in three variants: GWM Worlds for explorable environments, GWM Avatars for conversational characters, and GWM Robotics for robotic manipulation.

Why Is This Important? Runway is pivoting from "best video model" to "platform where you run all the models" — a smart hedge as the model competition keeps intensifying. For commercial shops, having Kling, Sora 2 Pro, and Gen-4.5 under one subscription and interface is genuinely workflow-relevant. The GWM-1 world model matters longer-term for environment and set generation, though it's clearly early-stage for production use. What to watch: whether the multi-model hub pricing holds up and whether GWM Avatars becomes a viable tool for digital human work.

What Happened: Lightricks has released LTX-2.3, framing the release as a philosophical shift: "For decades, creative software has been defined by its interface. We think the next era gets defined by the engine underneath." The update focuses on improved visual quality, better prompt adherence, and a suite of upscaler models that push generated content toward higher resolutions and smoother frame rates. The release was accompanied by a desktop video editor enabling the entire model to run locally on consumer hardware.

Why Is This Important? LTX-2.3 is the clearest signal yet that open-source video generation has arrived as a legitimate production option — not a hobbyist experiment. The local-run desktop app means IP-sensitive projects never leave the building, which matters enormously for pre-production on unreleased content. The 47GB model size is still a real hardware constraint for smaller shops, but the open weights and Apache 2.0 license mean the VFX community can fine-tune on house styles. This is worth a serious look, especially if your team is already running ComfyUI workflows.

What Happened: Beeble has launched SwitchX, an AI tool to change the lighting, background, props, costumes, and on-screen world of a video while keeping the subject consistent, recognizable, and intact. The workflow centers on ease of use: users upload a video, SwitchX automatically detects the main subject, and a reference image defines the desired scene including lighting, background, and props. The system analyzes footage frame by frame to maintain temporal consistency. Pricing starts at free for 5-second videos and scales to $60/month for 60-second clips at 2K resolution.

Why Is This Important? For commercial production, this directly addresses one of the most common client revision requests: "can we change the location/time of day/background?" without a reshoot. The pitch — keep the performance, switch everything else — is exactly what post houses need. The cloud-only limitation and questions around data security for sensitive footage are the friction points that will determine whether this gets adopted at the facility level vs. staying a solo-creator tool. Worth testing on a low-stakes job before trusting it on a :30 for a major brand.

What Happened: Amazon's biblical epic "House of David" featured 350 AI-generated shots in season two, a scale that would have been prohibitively expensive or time-consuming using traditional VFX methods. Amazon's AI Studio focuses on areas like improving character consistency across shots and supporting pre- and post-production, and the closed beta program launched this month with select industry partners; the company anticipates sharing initial outcomes by May.

Why Is This Important? Amazon is the first major studio to put an actual AI shot count on the record — 350 shots in a single series — which changes the conversation from "are studios using this?" to "how much and at what scale?" The closed beta this month is the test of whether these tools are production-grade or just good demos. The character consistency work is specifically the hard problem that's been limiting AI video for narrative content. Outcomes in May will be closely read by every post house pricing their VFX work.

What Happened: Artlist hosted a panel at SXSW 2026 bringing together practitioners to honestly assess where AI filmmaking stands. Josh noted that just a year ago, eight seconds of coherent footage felt like a ceiling — now it's fifteen. Tools like Kling 3.0 Motion Control are already pushing what's possible, maintaining visual consistency across angles, camera directions, and longer sequences. The panel also named a clear benchmark for current failure: a Thanksgiving dinner table scene with twelve people, overlapping dialogue, and a dog — realistic simultaneous action that AI still can't manage reliably.

Why Is This Important? This panel is worth reading in full because it's one of the more honest assessments of where the tech actually stands for working filmmakers — not hype, not doom. The "15 seconds of coherence" benchmark is a useful calibration. The gap between what AI generates and what a director with a specific vision actually needs remains real and significant, especially for narrative storytelling. The practical takeaway: AI is a serious tool for B-roll, VFX plates, and pre-viz — not yet for scripted, multi-character scenes. Plan accordingly.