In partnership with
Welcome to VP Land! We are back! And a happy 250th 4th of July to our American readers 🇺🇸
We've been busy with lots of new videos coming to YouTube soon, which overloaded our resources. But we're back! (and also..if you got an email about a new article on the website, sorry about that. It was meant to just go online, not emailed to everyone)
In today's edition:
Midjourney demands studios' internal AI files
Beeble rebuilds 16-bit HDR from 8-bit
Disney Imagineering taps Adobe Firefly
Open tools from LTX, Palmier, MiniMax
Midjourney Moves to Force Disney, Universal and Warner Bros. to Reveal How They Use AI
Midjourney is trying to turn its copyright fight with Hollywood into a look inside the studios' own AI work. In the infringement case Disney, NBCUniversal, and Warner Bros. Discovery brought over AI training, Midjourney filed a motion asking a federal judge to widen discovery so the studios must disclose how they use AI internally, Variety reports.
What Midjourney is demanding. The request reaches past the studios' consumer-facing products to internal AI business plans, research reports, training datasets and model weights, and board presentations about AI strategy.
The unclean-hands argument. "If Plaintiffs are doing the very thing they seek to punish, that evidence goes to the heart of Midjourney's fair use and unclean hands defenses," attorney Bobby Ghajar argued. If the studios train models on unlicensed material themselves, that speaks to industry custom and cuts against their own infringement claim. The studios' lead attorney, David Singer, has called the request a "fishing expedition" meant to distract from Midjourney's own conduct.
Where it stands. A magistrate judge had capped discovery at the studios' consumer-facing AI. Midjourney is asking District Judge John Kronstadt to overturn that limit and open up the internal use.
Why it matters. These suits have been framed as Hollywood versus AI, yet Disney is itself building internal AI tools, like the Imagineering park-design models below. A ruling for Midjourney would put that kind of use on the record and force the studios to defend their own AI practices, not just attack Midjourney's. The suits, which we covered when Midjourney launched its first video model, remain active.
SPONSOR MESSAGE
You've seen the AI demos. Viktor does it without you watching.
The AI tool you tried last quarter waited for a prompt, hallucinated a number, then asked if you'd like a summary.
Viktor opened a PR at 2am, rebased it against main, ran your test suite, and posted a note in #eng: "Two flaky tests in payments service, both pre-existing. Recommended merging after fixing them." Then drafted the customer reply for the support ticket the bug created.
That's 619K autonomous actions per day across 20,000+ teams. Not chat replies. Real work shipped to GitHub, Stripe, Linear, Notion, and 3,000+ other tools, from inside Slack and Microsoft Teams.
You don't supervise him any more than you supervise a senior engineer.
SOC 2 certified. Your data never trains models.
"It's what you probably originally thought AI was going to be when you first heard of it in sci-fi movies." Tyler, CEO.
Beeble's SwitchHDR Reconstructs True 16-Bit HDR From 8-Bit Video
Beeble released SwitchHDR, a model that reconstructs true 16-bit HDR from ordinary 8-bit SDR video. It rebuilds clipped highlights and pulls clean detail out of noisy shadows instead of tone-mapping a fake.
Scene-linear output, not a filter. SwitchHDR writes a true 16-bit EXR sequence in scene-linear ACES AP0 (ACES2065-1), stays temporally consistent across frames, and is controllable by region and prompt.
Aimed at the grade and VFX bays. It targets DITs, colorists, and VFX pipelines that need genuine dynamic-range recovery, and it complements Beeble Studio, the on-prem app we covered that runs SwitchLight 3.0 relighting at 4K on your own GPU.
Why it matters. HDR is now the delivery standard across streaming and premium theatrical, but most archives, stock, and lower-end camera output are still 8-bit SDR. A model that credibly reconstructs scene-linear HDR from that footage is a path to bring legacy and prosumer material up to current spec without a reshoot.
Disney Imagineering to use Adobe AI for theme park design
Walt Disney Imagineering is building custom Adobe Firefly models trained on its own park-design assets, according to BlogMickey.
What the models do. Early workflows include sketch-to-image concept art, on-brand franchise asset generation, and 2D-to-3D prototype tools for design iteration, materials planning, and engineering coordination.
Built on Firefly Foundry. Adobe's service trains a model on a single customer's own assets, so Imagineering's output stays inside Disney's IP rather than drawing on a general dataset. That is the pitch for a company that both owns valuable franchises and is suing over how AI is trained.
Why it matters. Disney is a lead plaintiff in the Midjourney case above, and this is exactly the internal, behind-the-camera AI use Midjourney wants exposed. We covered the company forming a dedicated unit to coordinate its AI and AR work; custom park-design models show how far that adoption already reaches.
MiniMax launches Hub all-in-one AI video generator at Shanghai Film Festival
The open-model drumbeat keeps rolling. Two of the tools below are open-source, one folds the whole pipeline into a single app, and Midjourney offered a rare technical look under its own hood.
MiniMax Hub. MiniMax launched Hub at the Shanghai International Film Festival, an all-in-one platform that bundles image, video, voiceover, music, and editing. You hand it a natural-language goal, a PDF, or reference assets, and an agent breaks the job into tasks, picks models, runs them, and checks quality, with human-in-the-loop checkpoints plus Skill and Memory functions that learn your workflow and aesthetic.
Midjourney publishes technical dive on new Midjourney Scanner
Midjourney Scanner. Midjourney's first hardware is a full-body ultrasonic scanner from its new Midjourney Medical division, not an image tool. It maps the body in under 60 seconds at MRI-like detail using a ring of roughly half a million ultrasonic elements, reusing the denoising math behind image diffusion. Body-composition maps come first, with a San Francisco location planned for 2027.
Palmier launches open-source Mac-native AI video editor
Palmier. Palmier.io is an open-source, Mac-native video editor that Claude can drive directly. You use AI to edit, organize, and generate footage right in the timeline, and it is free and available now. For editors already working in Claude, it puts an agent inside a real editing app instead of a chat window.
ByteDance launches Seedance 2.0 Mini for lower-cost AI video production
Seedance 2.0 Mini. ByteDance's Seedance 2.0 Mini, which we broke down, is a lower-cost tier of its multimodal video model that runs image, video, and audio references at about half the price of Seedance 2.0. It is built for everyday production work where you want fast iterations without paying full-model rates.
LTX launches open-source Trainer for custom video/audio LoRAs
LTX Trainer. LTX open-sourced LTX Trainer, a framework for training your own LoRAs and IC-LoRAs across video, audio, and cross-modal workflows. It ships with an agentic training skill, flexible conditioning, and free IC-LoRAs, with code, docs, and weights on GitHub and Hugging Face. If you want a model tuned to a specific look, character, or sound, you can build it yourself.
A24 shares Backrooms BTS on practical builds and Blender
A24 broke down the Backrooms blimps with director Kane Parsons, showing how the suburban set came together through practical build-outs and Blender. It is a concrete look at blending physical construction with 3D work on the creator-led horror feature we covered when A24 signed Parsons.
Stories, projects, and links that caught our attention from around the web:
🎬 Culver City folded film incentives into its budget, adding free production parking, suspended business taxes, and a FilmLA tour as Netflix and MGM+ shoot locally.
👓 Snap's Specs, its first standalone consumer AR glasses, arrive this fall at $2,195.
📱 An NCM and aTwist deal puts vertical microdrama previews in the pre-trailer slot in theaters, the phone-sized format's first grab at premium theatrical airtime.
🪄 The Ted team turned Seth MacFarlane into Bill Clinton with Deep Voodoo face tech and a forensic ZBrush sculpt after prosthetics and pure CGI failed.
📺 Instagram is testing horizontal video, episodic series, and live programming for TVs, and is expanding Instagram for TV to U.S. Samsung smart sets.
👾 YouTube's AI-slop crackdown is demonetizing faceless creator channels, pushing operators to hire cheap on-camera hosts from Fiverr and Upwork.
👔 Open Job Posts
Generative AI Supervisor - Milan, Italy 🆕
VFX Engine
Unreal Engine Animator - Vancouver, BC 🆕
Sony Imageworks
Unreal Engine Operator - Riyadh, Saudi Arabia
Pixomondo
AI Architect & Technical Strategist - Remote
Machine Learning Engineer, PyTorch - Remote
Cuebric
📆 Upcoming Events
Jul 19 to 23 🆕
SIGGRAPH 2026
Los Angeles, CA
Aug 4 to 6 🆕
Ai4 2026
Las Vegas, NV
Sept 11 to 14 🆕
IBC 2026
Amsterdam, Netherlands
Oct 12 to 16 🆕
VIEW Conference 2026
Torino, Italy
Oct 21 to 22 🆕
NAB Show New York 2026
New York, NY
View the full event calendar and submit your own events here.
Thanks for reading VP Land!
Thanks for reading VP Land!
Have a link to share or a story idea? Send it here.
Interested in reaching media industry professionals? Advertise with us.
