It was a week where AI filmmaking stopped being a tools story and became an industry story. Google handed A24 $75 million to co-build filmmaker tools with DeepMind. NVIDIA shipped a detector that can sniff out synthetic video in 22 milliseconds. And HeyGen made a two-host podcast show that literally builds itself. Here's what happened.
1. Google DeepMind hands A24 $75M to build filmmaking tools
The biggest story of the week isn't a model release — it's a $75 million research partnership between Google DeepMind and A24, the indie studio behind Everything Everywhere All at Once. A24 gets access to DeepMind's research and infrastructure; DeepMind researchers get to build workflows with actual filmmakers — not in a lab, but on real sets.
Here's what this deal is not: it's not an IP deal, not a data-training deal, and not a production mandate. A24's content library stays locked. It's a bet that the next generation of film tools will be shaped by the people who shoot films, not by model benchmarks.
"We believe the best way to develop tools that empower artists is to work directly with them. By collaborating with filmmakers and industry leaders like A24 from the beginning, we can build new AI features to support artists in authentic, meaningful storytelling."
— Demis Hassabis, DeepMind CEO, via Google's announcement
Per IndieWire, this is the first time DeepMind has partnered directly with a studio — previous collaborations were with individual filmmakers like Darren Aronofsky. A24's tech arm, A24 Labs (led by Scott Belsky), will be the bridge. If this works, it sets the template for how studios and labs collaborate: researchers build, filmmakers direct.
2. NVIDIA ships a synthetic-video detector at SIGGRAPH
At SIGGRAPH 2026 in Los Angeles, NVIDIA unveiled the Synthetic Video Detector — an AI NIM microservice that analyzes video frame by frame and scores how likely it is to be AI-generated. The numbers are serious: 92% accuracy on uncompressed video, 87% at 15% compression, 82% even at 50% compression — and it processes 1080p in about 22 milliseconds on RTX systems.
Why this matters to you: every AI filmmaker is going to meet this tool in the wild. Broadcasters, platforms, and brands are adopting detection exactly like this, and it's robust to the resizing, cropping, and re-encoding that happen in every social pipeline. Knowing how your output reads to a detector — compression levels, re-encoding artifacts, over-smoothing — is now part of the job.
NVIDIA also opened up Cosmos 3 Edge, its compact world model, for local edge GPUs, and pushed MCP connections deep into creative tools: Adobe, Blender, Houdini 22, Unreal Engine, Foundry, and Boris FX all showed AI agents that can inspect scenes, flag issues, and prep exports inside the app. The agentic assistant is coming to the NLE — probably sooner than the NLE vendors are ready for.
3. HeyGen turns any link into a two-host video show
HeyGen's new Video Podcast tool takes a URL, PDF, or a bare idea and produces a complete two-host podcast episode: AI-written dialogue, assigned speaker roles, studio scenes, multi-camera cuts, B-roll, captions, and chapters. No mics, no recording, no edit. Ready in minutes.
Reactions on X split exactly down the middle — half the replies asking why anyone would watch an AI podcast, the other half noting that some AI episodes already out-produce the human filler clogging the feed. Either way, the workflow is a genuine time-saver for repurposing long-form content into a video-native format. It's not a filmmaking tool, but it's the fastest path to a publishable video show from a document that currently exists.
4. Runway's AI short films get taken seriously
Runway unveiled the ten finalists of its AI short-film program, and Fast Company's verdict was blunt: the medium has "come far enough to be taken seriously." Ten films, ten different directors, all generated with AI tools — and the conversation has moved past "is this real?" to "is this good?"
That's the shift worth watching. When the finalists of an AI film program are judged on craft instead of novelty, the tools have crossed a threshold. The next step is distribution — and that's exactly where the A24 deal comes back in.
The takeaway
One week, four stories, one direction: AI filmmaking is consolidating from "which model do I prompt?" into "what's the production pipeline?" Studios are signing research deals, platforms are building detection, and the tools are getting opinionated about workflow. The filmmakers who win the next 12 months won't be the ones with the best prompts — they'll be the ones with the best process.
Links: Google × A24 (Variety) | NVIDIA at SIGGRAPH | HeyGen Video Podcast | IndieWire analysis
GenFlick is building the AI filmmaking stack.
Join 2,000+ filmmakers and creators who read this newsletter every week.
Sign Up Free →