Welcome to Today’s AIography!

The AI filmmaking landscape is moving in two directions at once. Luma's new Agents are the first true creative collaborators that plan and execute across text, image, video, and audio without constant prompting. Peter Diamandis just put $3.5M on the table for optimistic sci-fi, Canal+ is integrating Veo 3 into production pipelines, and NVIDIA's making local generation accessible at GDC. But the most powerful AI video tool ever built—Seedance 2.0—is locked behind “The Great Firewall of China” access and indefinite legal delays, creating a gold rush of scam sites claiming they can get you in early. They can't. Today's issue covers the tools you can use, the competition worth entering, and the scam landscape you need to avoid. The shift isn't incremental anymore—it's architectural. So is the chaos.

 In today’s AIography:

• Luma Agents: End-to-end creative AI powered by Unified Intelligence

• Future Vision XPRIZE: $3.5M for optimistic sci-fi films

• Why You Can’t Find a Real Version of Seedance 2.0

• NVIDIA & ComfyUI streamline local AI video at GDC 2026

• Canal+ partners with Google Veo 3 for production workflows

• Essential Tools

• Short Takes

• One More Thing…

Read time: About 8 minutes

THE LATEST NEWS

VIDEO GEN
Luma Agents: Creative AI That Thinks, Plans, and Executes End-to-End

TL;DR: Luma launched Luma Agents on March 5—AI agents powered by its new Unified Intelligence (Uni-1 model) that handle complete creative workflows across text, image, video, and audio, coordinating with Ray 3.14, Veo 3, Seedream, and ElevenLabs.

Key Takeaways:

• Uni-1 model: First unified multimodal AI trained on audio, video, image, language, and spatial reasoning in a single system—what CEO Amit Jain calls "intelligence in pixels"

• Agentic workflow: Plans, generates variations, self-critiques, and iterates without constant prompting—similar to coding agents' "check your work" capability

• Already deployed: Publicis Groupe, Serviceplan, Adidas, Mazda, and Saudi AI company Humain are using it in production

• Real-world results: Turned a $15M, year-long ad campaign into localized ads for multiple countries in 40 hours for under $20K, passing brand QC

• Available now via API with gradual public rollout to ensure reliability

Why It's Important:
This isn't another model you prompt—it's a shift in how AI creative work gets done. Instead of bouncing between 10 tools and iterating manually on every asset, Luma Agents maintains persistent context across your project, generates large variation sets, and lets you steer through conversation. Jain's right: "Our customers aren't buying the tool; they're redoing how business is done."

For filmmakers and agencies, this is the pivot point where AI stops being a render farm and starts being a production assistant that actually understands your brief. The Unified Intelligence architecture—thinking in language and rendering in pixels simultaneously—is what makes this possible. It's the same mental model a human DP uses when framing a shot: understanding intention, space, light, and execution at the same time.

If you've been waiting for AI to feel less like a complicated prompt lottery and more like a collaborator, this is the first real answer.

FILM INDUSTRY
Future Vision XPRIZE: $3.5M Prize for Optimistic Sci-Fi Films

TL;DR: XPRIZE founder Peter Diamandis announced the Future Vision XPRIZE on March 9—a $3.5M+ global competition asking creators to imagine abundant, technology-driven futures for humanity using AI filmmaking tools.

Key Takeaways:

• $3.5M+ total prizes for optimistic sci-fi short films and features

• Focus: Abundant futures powered by technology (opposite of dystopian narratives)

• Open to global creators using AI video generation tools

• Announced March 9, 2026 by Peter Diamandis (founder of XPRIZE Foundation)

• Competition details and submission timelines TBA

Why It's Important:
Diamandis is putting serious money behind a narrative shift. While Hollywood churns out AI-apocalypse stories, XPRIZE is funding creators to imagine futures where technology solves problems instead of creating them. This competition legitimizes AI filmmaking at a scale that matters—XPRIZE competitions have historically driven innovation in space travel, ocean exploration, and carbon capture.

For AI filmmakers, this is your shot at funding, visibility, and validation from one of the most respected innovation organizations on the planet. The brief is wide open: abundant futures. That could be post-scarcity economies, AI-assisted medicine breakthroughs, climate restoration, or space colonization. The constraint is optimism—a rare requirement in sci-fi.

If you've been building AI filmmaking skills, this competition is your runway. Start ideating now.

VIDEO GEN
Why You Can’t Find a Real Version of Seedance 2.0

TL;DR: ByteDance's Seedance 2.0 is arguably the most powerful AI video tool ever built — and you can't have it. It launched February 12, China only, and the global API release has been indefinitely postponed under Hollywood legal pressure. That gap between massive hype and zero access has created a gold rush of third-party websites claiming they already have the model. They almost certainly don't. Here's what you need to know before you hand over your credit card.

First, Why Everyone Wants It

Before we get into the scam landscape, you need to understand why the demand is real enough to be exploited.

Irish director Ruairi Robinson used Seedance 2.0 to generate a 15-second clip of Tom Cruise and Brad Pitt fighting on a crumbling rooftop at twilight — sweeping camera angles, stunt choreography, sound design, music. All from two sentences. That single clip triggered cease-and-desist letters from Disney, Warner Bros., Paramount, Sony, and Netflix, who collectively labeled it a "high-speed piracy engine."

ByteDance describes Seedance 2.0 as giving creators "director-level control" over performance, lighting, shadow, and camera movement — supporting up to 9 image references, 3 video clips, and 3 audio files in a single generation. Tech analysts are calling it the "second DeepSeek shock," a model that rivals Western tools like OpenAI's Sora 2 at a fraction of the cost.

It's real, it's extraordinary, and almost nobody outside China can touch it.

Why You Can't Have It (Yet)

The official launch on February 12 was strictly for mainland China via the Jimeng AI platform — access requires Chinese phone verification, effectively locking out international users. A global API release planned for February 24 was indefinitely postponed as ByteDance scrambles to implement IP filters and identity protection mechanisms under intense legal pressure. No new date has been announced.

That's the gap. Massive hype. Zero legitimate international access. A window of confusion that bad actors are walking straight through.

The "Buyer Beware" Part — Which Is Really Why We're Here

In the weeks since launch, a cottage industry of third-party websites has emerged, all claiming to offer Seedance 2.0 access to international users. They have professional-looking interfaces, testimonials, pricing tiers, and the Seedance name front and center. Some even rank well in search results.

Here's what's actually going on under the hood.

BytePlus — ByteDance's own enterprise division — has explicitly stated that "various external API services are fake," and that the only possible implementation method for any third-party claiming Seedance 2.0 access is reverse-engineering ByteDance's Dreamina web version. That's not a loophole. That's unauthorized access to a platform you didn't build, resold to you at a markup.

Take seedance2.ai as an example. It markets itself as a Seedance 2.0 platform, complete with that branding throughout. But look at its own model selector — it reads Seedance 1.5 Pro. Not 2.0. The site's own interface exposes the gap between what it promises and what it delivers. That's not a bug. That's the business model.

When you use one of these sites, here's what you're actually signing up for:

• You don't know what model you're running. It may be 1.5 Pro, a different model entirely, or something reverse-engineered with no quality guarantees.

• Your data has zero protection. There is no official relationship with ByteDance, no terms of service with any meaningful legal backing, and no visibility into what happens to your prompts, your reference images, or your generated content.

• You could be funding account abuse. The reverse-engineering approach works by simulating user logins to Dreamina and intercepting communication protocols — posing risks of account bans, data leaks, and legal exposure for everyone in the chain.

• The "testimonials" mean nothing. Stock photo avatars with generic praise are a five-minute job. Several of these sites share the same boilerplate review structure.

How to Spot These Sites

A few red flags to watch for:

• Claims of "Seedance 2.0 access" with no mention of China-only restrictions

• Pricing in USD with immediate access — no waitlist, no verification friction

• Testimonials with Unsplash-style profile photos

• No physical address, no company name, no real support contact

• Model selector or UI that quietly reveals a different (older) model

• Aggressive SEO copy built around "how to access Seedance 2.0 outside China"

What to Do Instead

Wait for the official release — and watch for it at the only legitimate source: seed.bytedance.com. ByteDance will announce global API availability through official channels. When it happens, you'll know.

In the meantime, Veo 3.1 and Kling are both legitimate, available, and genuinely capable tools. You won't be sitting idle — you'll be building workflow skills that transfer directly to Seedance when it opens up.

The model is worth waiting for. The scam sites are not worth the risk.

VIDEO GEN
NVIDIA and ComfyUI Streamline Local AI Video Generation at GDC

TL;DR: NVIDIA announced at Game Developers Conference (March 11) major updates for local AI video generation on RTX GPUs, including ComfyUI's new simplified App View, RTX Video Super Resolution, and new NVFP4 models.

Key Takeaways:

• ComfyUI App View: Simplified interface for node-based AI video workflows—lowers barrier to entry for game devs and creators

• RTX Video Super Resolution: Enhanced upscaling for AI-generated video on NVIDIA RTX GPUs

• NVFP4 models: New optimized models for RTX hardware

• Local generation focus: Run AI video tools on your own hardware without cloud dependency

• GDC 2026 positioning: Aimed at game developers, but applicable to all creators

Why It's Important:
Cloud AI video tools are convenient until they're not—rate limits, subscription costs, data privacy concerns, and internet dependency. NVIDIA's doubling down on local generation means creators can own their entire pipeline. ComfyUI's App View is the key unlock here: it turns node-based workflows (intimidating for non-technical users) into accessible interfaces.

For game developers building cinematics, this is native integration. For filmmakers with RTX rigs, it's creative freedom without monthly fees. As cloud tools consolidate and prices rise, local generation becomes the hedge against platform lock-in.

If you own an RTX GPU, this update just made your hardware significantly more valuable..

FILM INDUSTRY
Canal+ Partners with Google Veo 3 for AI-Driven Production

TL;DR: French broadcaster Canal+ announced March 11 a partnership with Google to provide Veo 3 video AI to production teams for pre-visualization, historical recreation from archival photos, and AI-powered content recommendation.

Key Takeaways:

• Pre-visualization workflows: Directors can prototype scenes before shooting

• Historical recreation: Generate video from archival photographs (documentaries, historical dramas)

• Content recommendation: Google AI powers viewer recommendations across Canal+ platforms

• Production team deployment: Available to Canal+ creators and partners

Why It's Important:
This is institutional validation for AI video in traditional broadcast workflows. Canal+ isn't experimenting—they're integrating Veo 3 into production pipelines for directors and crews. Pre-vis has always been expensive and time-consuming; AI generation makes it instant and iterative. Historical recreation from photos unlocks documentary storytelling that was previously impossible or prohibitively expensive.

For filmmakers, this signals the path forward: major broadcasters are adopting AI tools as production infrastructure, not gimmicks. If you're pitching projects to broadcasters or production companies, knowing these tools and workflows is rapidly becoming table stakes.

The Canal+ deal also shows Google's enterprise strategy: embed Veo 3 into established production environments where reliability and integration matter more than bleeding-edge features.

AI Filmmaking & Content Creation Tools Database

Check out the Alpha version of our AI Tools Database. We will be adding to it on a regular basis.

Got a tip about a great new tool? Send it along to us at: [email protected]

SHORT TAKES

• Sora Launches Android App with Extended Features

• Tropfest AI Film Backlash Continues

• Seedance 2.0 vs Kling 3.0: Benchmark Comparisons Emerge

• World AI Film Festival (WAIFF) 2026 Lineup Announced

• "20 Seconds for Art" Competition: Humans and Intelligent Machines

• Kling 3.0 vs Seedance Positioning Debate Heats Up

• Runway Characters Launches Real-Time AI Avatars
  

ONE MORE THING…

Video of the Week

"Seedance 2.0 — Best AI Video Generations Part 3 (It Keeps Getting Better)"

A curated compilation of 30+ standout AI-generated videos from Seedance 2.0's multi-input system—showcasing character consistency, motion referencing, and cinematic camera work that feels directed, not randomly generated. This is the best public showcase of what reference-based AI video generation can do when you feed it visual language instead of text prompts.

If you're skeptical about AI video quality, this compilation will change your mind. The consistency across shots, the coherent motion, and the cinematic framing prove that AI video has crossed the "looks good enough to use professionally" threshold.

Final Thoughts

The pattern is clear: reference beats text. Seedance's multi-input system, Luma's Unified Intelligence, NVIDIA's local workflows—they all point toward the same evolution. The next generation of AI video tools won't ask you to describe what you want in words. They'll ask you to show them.

This is how filmmakers actually work. You don't write a paragraph describing a dolly shot—you show your DP a reference. You don't describe your character in prose—you pull a look book. AI tools are finally catching up to how visual storytelling actually happens.

The winners in the next 12 months will be the creators who stop trying to "prompt better" and start building visual reference libraries, motion banks, and style guides. The tools are here. The workflow is shifting. The question is whether you're shifting with it.

Stay sharp,
Larry

P.S. If you're entering the Future Vision XPRIZE competition, I want to hear about your project. Reply with your concept—optimistic sci-fi built with AI tools is exactly the kind of work this community should be championing.

AIography is written by Lawrence Jordan, who has decades as a Hollywood editor (Emmy & ACE Eddie nominated), founder of Lumarka AI filmmaking platform, and your guide through the AI production revolution. Every issue distills the signal from the noise so you can build, ship, and stay ahead.

Want more? Join 800+ AI filmmakers in the AIography Skool Community for daily discussions, workflow breakdowns, and early access to tools. Founding Members ($29/month) get exclusive courses and direct access.

Share this newsletter: Forward to a filmmaker friend who needs to see where this is headed.

If you have specific feedback or anything interesting you’d like to share, please let us know by replying to this email.

AIography may earn a commission for products purchased through some links in this newsletter. This doesn't affect our editorial independence or influence our recommendations—we're just keeping the AI lights on!