The week of March 30 – April 5, 2026 crystallized one of the biggest power shifts in AI video to date. With Sora’s shutdown now official and its user base scattering to competitors, the rest of the industry moved fast — Google launched a budget-friendly video API, Runway bet $10 million on a new startup ecosystem, and SAG-AFTRA floated a radical tax proposal that could reshape how studios think about synthetic performers.

The Post-Sora Scramble Is Now a Data Story

Bloomberg published new numbers on April 1 confirming what many suspected: the AI video market is absorbing Sora’s users at speed. Kling AI, owned by China’s Kuaishou Technology, saw global weekly active users climb 4% to an average of 2.6 million in the week following OpenAI’s announcement. Even before the shutdown news, Kling was already leading globally with 7.8 million monthly active users in March compared to Sora’s 4.7 million. RunwayML and Vidu each posted a more modest 1% uptick in weekly active users over the same period. The Sora website and standalone app will go dark on April 26, with the developer platform following on September 24. OpenAI is keeping limited video generation capabilities inside ChatGPT for Plus and Pro subscribers, but the era of Sora as a standalone product is effectively over.

Google Launches Veo 3.1 Lite — The Cheapest Frontier Video API Yet

On March 31, Google released Veo 3.1 Lite into the Gemini API, positioning it as the most cost-effective frontier video generation model currently available to developers. At $0.05 per second for 720p video and $0.08 per second for 1080p, Veo 3.1 Lite runs at less than half the cost of Veo 3.1 Fast while matching its generation speed. The model supports both text-to-video and image-to-video workflows, with flexible aspect ratios for landscape and portrait formats and generation lengths of 4, 6, or 8 seconds. Google also signaled that pricing for Veo 3.1 Fast will drop further in April. The move reads as a clear land-grab: with Sora gone and Runway focused on premium creator tools, Google is betting that volume pricing will pull developers building marketing, e-commerce, and social media tools onto its platform.

Runway Bets $10 Million on the AI Video Builder Ecosystem

In a complementary but different strategic play, Runway announced a $10 million fund and a new Builders program on March 31. The fund, seeded by existing investors and close partners, will write checks of up to $500,000 for pre-seed and seed-stage startups building on top of Runway’s models. The Builders program extends to seed through Series C companies, offering 500,000 free API credits and early access to Characters — Runway’s real-time video agent API powered by its general world models. Runway is targeting three categories of builders: teams pushing the frontier of AI architecture, companies creating the application layer on top of foundation models, and ventures experimenting with new forms of media creation and storytelling. At a $5.3 billion valuation following its $315 million raise in February, Runway is signaling that it sees its future less as a standalone tool and more as infrastructure for an entire ecosystem.

SAG-AFTRA’s “Tilly Tax” Could Make AI Actors as Expensive as Real Ones

The most consequential story for filmmakers this week may have come from the bargaining table. Reports from Bloomberg and Variety revealed that SAG-AFTRA is pushing a so-called “Tilly tax” — named after controversial AI actress Tilly Norwood — in its ongoing contract negotiations with Hollywood studios. The proposal would levy a royalty fee on studios every time they use a fully synthetic AI-generated performer, with the explicit goal of making AI actors cost as much as hiring real ones. The revenue would flow back to the guild and be distributed to working performers whose livelihoods are most threatened by AI replacement. With both the DGA and SAG-AFTRA contracts expiring on June 30, and newly elected SAG-AFTRA president Sean Astin leading alongside chief negotiator Duncan Crabtree-Ireland, these talks carry real stakes. If the Tilly tax makes it into the final contract, it would be the first financial mechanism anywhere in the world designed to equalize the cost of synthetic and human screen talent.

ByteDance’s Seedance 2.0 Faces Infringement Backlash

ByteDance continued its aggressive push into AI video with Seedance 2.0, now integrated into CapCut and available via the Dreamina platform. The model represents a serious technical step up, leaning into art-direction workflows and positioning itself as a direct competitor to Kling and Veo in the high-end generation market. But the rollout has not been smooth. SAG-AFTRA publicly condemned ByteDance earlier this year for what it called “blatant infringement,” citing Seedance-generated videos that included unauthorized use of the voices and likenesses of members including Tom Cruise and Brad Pitt. The controversy underscores a tension that is only growing: as these models get better at generating realistic human performances, the legal and ethical questions around training data and output similarity become harder to ignore.

The Free AI Video Era Is Officially Over

One broader trend crystallized this week: the free tier of AI video generation is disappearing. With Sora gone, Grok Imagine moving to paid-only, and Google’s Veo 3.1 Lite designed as an affordable-but-not-free developer tool, the market has consolidated around premium pricing. The platforms that survived — Runway, Kling, Veo, and Pika — did so by building products that users will pay for. For independent filmmakers and small studios, this means budgeting for AI video tools is no longer optional; it is a production cost, just like camera rental or location fees.

Open-Source Video Models Keep Closing the Gap

While the commercial landscape consolidates, the open-source community continues to make gains. Wan 2.2 remains widely regarded as one of the most cinematic open-source video generation models available, using a Mixture-of-Experts diffusion backbone that distributes denoising across specialized expert networks. HunyuanVideo, operating in a spatial-temporal latent space with a dual-stream transformer architecture, continues to develop as a viable alternative for researchers and developers who need flexibility without API costs. Lightricks’ LTX 2.3, a 22-billion-parameter model released in early March, has shown that open-source can now generate 4K output — territory that was exclusively commercial just six months ago.

What March 30 – April 5 Tells Us About Where AI Video Is Heading

This was a week that drew clear battle lines. Google is competing on price. Runway is competing on ecosystem. SAG-AFTRA is competing on policy. And the open-source community is competing on access. The Sora shutdown removed the most prominent name in AI video but did nothing to slow the field — if anything, it accelerated the moves that its competitors were already making. For filmmakers, the practical takeaway is that the tools are better, cheaper, and more numerous than ever, but the rules around how they can be used — especially with human likenesses — are tightening fast. The Tilly tax debate will be one to watch closely over the next three months as contract deadlines approach.