IATSE hires its first tech officer for AI

Sponsored by

Welcome to VP Land! It's been a minute...but we are back! And upgraded. A few things that are new on the website: new tool directory, better job board, and upgraded events page. Plus the latest jobs, events, and tools will get highlighted in each newsletter.

AI is reaching the bargaining table, the review pages and live broadcasts, with a crew union hiring a dedicated AI officer, critics weighing in on a generated epic and Prime Video rebuilding real plays in 3D.

In today's edition:

IATSE hires its first labor and technology intelligence officer
Critics pan Grok's AI-generated "Odyssey" episode
Prime 360 rebuilds NBA replays from new angles
Kling 4.0 is out, plus FLUX 3 Max and OmniHuman 2.0
Tencent's open-source world model

IATSE Names Jillian Arnold Its First Labor and Technology Intelligence Officer to Cover AI

IATSE hired former Local 695 president Jillian Arnold as its first labor and technology intelligence officer, a role that covers AI in contract talks, member reskilling and policy. Arnold says the goal is crew helping shape how AI enters their crafts.

According to The Hollywood Reporter, Arnold will work on contract proposals, training, and state and federal AI legislation for a union of more than 170,000 members across the U.S. and Canada. She said the union wants to be "a part of the deployment and integration of any AI within our craft so that we make tools that are for workers, not displacing workers."

A crew background. Arnold is a video engineer and former cinematographer whose credits include NASA's Mars Curiosity Rover landing and live broadcasts of the Oscars and the Emmys. In 2020 she became the first woman and first video engineer elected president of IATSE Local 695.
She chaired the union's first AI subcommittee. IATSE formed its AI and emerging technologies subcommittee in 2023 with Arnold as chair, and she was one of the negotiators in the 2024 talks with the AMPTP.
Contract proposals come first. Her top priority is proposals for the Basic and ASA agreements ahead of the next negotiations, alongside state legislation and federal work with IATSE's political department. Per TheWrap, she will bridge technology's growing role with the union's commitment that productions "remain authentically made by humans."
A case for analog workflows. A 2024 consultation provision lets members discuss AI use with an employer before employment begins. Arnold said crew can use it to argue for analog workflows that produce "a unique, copyrightable solution."
No model endorsements. Asked about ethically trained models, Arnold said IATSE is not endorsing any model. The union has also rolled out a class teaching Local leaders how to talk to members about AI.

Arnold said leadership has identified "first wave" crafts it is engaging with, but declined to name them without data to back it up; the bargaining approach will take shape through committee work on member proposals. We covered Arnold's role and priorities in more depth.

SPONSOR MESSAGE

State of Product reveals what AI still hasn’t solved

80% of product professionals say AI helps them ship faster, but customers aren’t seeing value any sooner. Atlassian’s State of Product 2027 explores the gains AI is delivering and the challenges that remain, from decision-making that hasn’t kept pace to gut instinct overriding customer evidence.

Read the report

Critics Pan Grok's First AI-Generated 'Odyssey' Episode

Reviewers and viewers are panning the first seven-minute episode of Grok Imagine's AI-generated "Odyssey" series, calling out generic faces, flat dialogue and trailer-style pacing. The episode is a partial delivery on Elon Musk's promise of a full-length film.

Grok Imagine posted Episode 1 on X, crediting the "creative collaboration" to Wonder Studios and director Dev Walker, as Variety reports. Musk had promised a "historically accurate" feature-length version of Homer's poem by the end of 2026, after repeatedly criticizing the casting of Christopher Nolan's "The Odyssey."

Trailer shots with no story. Digital Trends called it "seven minutes of nothing happening beautifully" and wrote that it "feels like an extended trailer for a film that doesn't exist." Its strongest shots, the review says, are landscapes without actors.
Faces and voices give it away. The same review says Penelope has "the familiar beauty of an Instagram filter asked to look worried," Odysseus is hard to tell apart from other dark-haired men, and every line lands "robotically."
"Pure garbage" online. Euronews collected reactions calling the clip "pure garbage," with one user noting that a project billed as historically accurate contains dozens of historical inaccuracies.
The kindest notice. TechCrunch wrote it was "not as spectacularly ugly as I'd expected," with moments that could pass for real actors on a real sea, while noting that no one speaks ancient Greek. The episode credits Wonder Studios, which specializes in AI filmmaking.

The reviews single out performance, faces and voice as the weak spots in generated footage. Grok Imagine has offered a full breakdown of how it was made if viewers ask for one, and TechCrunch's report has the full episode.

Prime 360 Is Prime Video's New Virtual Camera System for 3D-Reconstructed NBA Replays

Prime Video introduced Prime 360, a virtual camera system that rebuilds recorded NBA action as a 3D scene so replays can show angles no physical camera captured. Every replay starts from footage of a real possession.

Sports on Prime announced the system as AI-powered technology that uses an array of high-res cameras around the court and AWS Trainium chips. It appears on select broadcasts within Prime Video's second NBA season, a slate of more than 100 games.

Perimeter cameras feed a cloud reconstruction. Prime Video's Woody Arrington told Sports Video Group that calibrated, synchronized cameras surround the court, and the capture moves to the AWS cloud to reconstruct the 3D scene.
Recorded plays, new viewpoints. Angles where no camera stood come from the reconstructed scene rather than a text prompt. Associate producer Josh Green called it "a replay machine" with "no plays that are off-limits."
Key specs are undisclosed. Prime Video has not given a camera count, reconstruction accuracy figures or processing times.

For virtual production crews, Prime 360 is a broadcast-scale example of capturing once and choosing the camera afterward. We covered how the Prime 360 pipeline works in more depth.

👨🏻‍💻 Takehound. Takehound is an AI video editor in the browser for long-form videos where you talk to camera.

💬 Scanfile. Translates text in existing SRT, VTT and ASS subtitle files, plus PDF, EPUB and PowerPoint (PPTX/PPT) documents, in a browser.

🔗 Treza. Treza is a node editor for AI video pipelines.

Browse the full tool directory

Kling 4.0 Leads a Batch of New Video, World and Vector Tools

Eight tool updates, sorted by where they land in a production: video generation, world and 3D building, and graphics.

Video generation

Kling 4.0 is out. Kling AI's new video model is fully available through a refreshed interface on kling.ai and through MCP and CLI, which Kling pitches for agent-driven batch work. 10-bit HDR output at 4K and 1080p and enterprise API access are listed as coming soon.
FLUX 3 Max. fal's post-trained video model, built with Black Forest Labs on FLUX 3, returns a 10-second clip in about 12 seconds, and fal reports a preliminary tie for #1 in Artificial Analysis's image-to-video arena. 720p output costs $0.0425 per second until November 7, 75% off the $0.17 list price.
OmniHuman 2.0. BytePlus's avatar model turns one image and an audio track into up to five minutes of performed video in a single continuous shot at native 720p, up from 60 seconds. The price drops to $0.07 per second from $0.12, so a full five-minute take costs $21.
Seedance 2.5 4K. BytePlus says its new 4K Seedance model brings more nuanced facial expressions and richer detail, with a Draft mode for testing shots as lower-cost 480p previews before rendering the final in 4K. Access is limited to selected enterprise customers, with API timing unannounced.

World and 3D building

Hy-Nomas. Tencent's new open-source video world model builds a scene segment by segment as you direct each turn with text or camera trajectories, with 720p output up to 60 seconds and code and weights under Apache-2.0. Default inference uses eight GPUs, while Tencent says a single RTX 4090 can run a low-memory configuration.
fal Worlds. fal's new platform puts interactive world models, including H3 Max Director, Lingbot World 2 and Abot World, on one page where you steer live-generated video as it streams. Public H3 Max Director sessions cap at two minutes, and fal describes the output as streamed video, with no mention of exported 3D geometry.
Viggle JST-2. Viggle's world model generates explorable 3D Gaussian splat scenes, characters from a single image, and full-body motion that exports as GLB or FBX for Mixamo and MetaHuman rigs. Viggle claims scene inference takes about a second, though scene collisions and object interactions still use rule-based physics; API pricing is $0.06 per prompted scene and $0.25 per character.

Graphics

Ideogram Vector. Ideogram's first vector model generates editable SVGs from a prompt or converts an existing image, so logos, titles and print graphics scale and recolor shape by shape. API pricing runs $0.13 to $0.32 per generated SVG and $0.10 per converted image.

I Solved Green Screen, and I'm Giving it to YOU!

Corridor Crew's 29-minute video walks through CorridorKey 2.0, the team's AI keying tool, from neural-network development to stress tests and feedback from professional VFX supervisors.

It's worth the runtime for anyone who pulls keys, since it shows how the tool was built, how it holds up under pressure and what working supervisors make of it. A free version is on GitHub, and the video is sponsored by NVIDIA and Puget Systems.

Stories, projects, and links that caught our attention from around the web:

📈 Anthropic made its US IPO filing public, disclosing rapid revenue growth and steep losses ahead of a planned Nasdaq listing under ANT, a look at what building frontier AI costs.

🛡️ WGA East members get free access to Cloak, which makes registered scripts resistant to AI scraping while keeping their formatting, so writers can send drafts out without feeding training data.

💰 Nvidia and TCV invested in Higgsfield, extending the AI video generator's Series B after it said it reached $1B in annualized revenue.

🕷️ Nicolas Cage said he refused to grant AI rights for Spider-Noir and may not work with Amazon again, putting AI contract language in public view.

🎭 Runway CEO Cristóbal Valenzuela said AI replacing actors is "already happening in a way," sparking View Conference's biggest debate as Nvidia's Richard Kerris pushed back.

We Debated Whether Vibe-Coded Clones Can Replace Pro Software

In Denoised, we ask what happens to software when AI can reverse engineer an app and rebuild it, from the REA tool and the vibe-coded Crafting Apps suite to Gemini 4 Argon and Opus 5.5 agents building detailed Blender scenes. Watch the episode for the demos.

We also debate where software companies keep their moat once code gets cheap, why the cloned video editor reportedly stutters on real-time previews and transitions, and what Meta's Muse Wonder is built for. Our episode recap has timestamps and notes.

Watch/Listen & Subscribe

Spotify | Apple Podcast | YouTube