Beeble SwitchX 3.0 levels up AI hybrid

Sponsored by

Welcome to VP Land! AI is moving closer to the shot itself, from Beeble directing crowds around a locked performance to CAA's scanned stars feeding Ogilvy ads, while Skydance's CEO pitches AI as the biggest creative unlock since CGI.

In today's edition:

Beeble's SwitchX 3.0 takes up to four reference images to guide environments
David Ellison calls AI the biggest creative unlock since CGI
Ogilvy and CAA launch Echo for talent-approved AI ads
ElevenLabs Scribe v3, ComfyUI ProRes export and Flow's 15-second shots
Blackmagic's 108-capsule ambisonic mic

Beeble Launches SwitchX 3.0, Built on MiniMax H3.0, With Up to Four Reference Images to Steer Environments

Beeble's SwitchX 3.0 lets artists guide a generated environment with up to four visual references, mixing reference images, floor plans and storyboards, and is built on MiniMax H3.0. Beeble says the original live-action performance stays intact frame by frame while the world around it changes.

Beeble announced the SwitchX 3.0 update to its video-to-video model with four main additions: multi-image references, dynamic backgrounds, improved camera tracking and more precise prompt control. It is available globally through the Beeble web app and Canvas.

Four references instead of one. The SwitchX 2.0 workflow paired a source video and a mask with a single reference frame. Version 3.0 accepts up to four, and artists can combine reference images, floor plans and storyboards to define complex spaces.
One green-screen take, three locations. In Beeble's demonstration, a green-screen performance becomes a continuous journey through a kitchen, a corridor and a rooftop, guided by four visual references, as Broadcast reported.
Built on MiniMax H3.0. Beeble built SwitchX 3.0 on MiniMax H3.0, then re-engineered it with its own dataset and architecture for professional video-to-video work. Output reaches 4K with 10-bit color and keeps the source footage's frame count and aspect ratio.
Environments that hold through camera moves. According to Beeble, integrated camera tracking keeps generated environments aligned with handheld, orbiting and running shots without tracking markers or a separate matchmove pass.
Background action is now promptable. Dynamic backgrounds can add moving crowds, and prompt control can trigger specific events such as birds taking flight as an actor passes. Hoon Kim, founder and CEO of AI VFX platform Beeble, said the original performance "stays intact, giving artists the consistency and control they need for professional production."

SwitchX 3.0 is included in all paid plans with no change to credit pricing, and 4K and 10-bit MOV downloads are limited to Professional and Team subscribers. We covered SwitchX 3.0 in more depth, with Beeble's demo frames.

SPONSOR MESSAGE

How 2M+ Professionals Stay Ahead on AI

AI is moving fast and most people are falling behind.

The Rundown AI keeps you ahead of the curve.

It's a free AI newsletter that keeps you up-to-date on the latest AI news, and teaches you how to apply it in just 5 minutes a day.

Plus, complete the quiz after signing up and they’ll recommend the best AI tools, guides, and courses — tailored to your needs.

Sign up to start learning.

David Ellison Calls AI the Biggest Creative Unlock Since CGI on His First Earnings Call Running Paramount and Warner Bros.

On his first earnings call running both Paramount and Warner Bros., Skydance CEO David Ellison said AI will help the company's studios make content faster and cheaper, saying AI "has the potential to be the biggest creative unlock in modern history." He paired that with a promise that "in a world of AI slop we think there will always be a premium for truly hand-crafted storytelling."

As The Hollywood Reporter reported, Ellison compared AI to the arrival of CGI and digital composites in the early '90s, which produced Jurassic Park and Terminator 2. Skydance plans to spend $30 billion a year across its content, from HBO prestige series and CBS procedurals to NFL rights and features at Warners and Paramount.

The Terminator math. Ellison said James Cameron made Terminator for $4 million in 1983, and a shot-for-shot remake would cost $60 million to $70 million today. He said AI tools could make budgets like that "more digestible" so studios take more risks on original stories.
AI sits inside the savings plan. Skydance is targeting at least $6 billion a year in savings within three years, and the shareholder letter lists AI workflows and AI coding tools among the ways to get there, The Next Web reports.
Streamers merge on an AI timeline. Ellison called merging Paramount+ and HBO Max within a year an "ambitious target" that is "possible because of the latest tools in artificial intelligence." A bundle of the two launches in early 2027, per TheWrap.
Franchises as the moat. Ellison described 40 franchises worth $1 billion each as protection against AI content, serving more than 203 million subscribers across Paramount+, HBO and HBO Max. He also said AI "will never be a replacement" for artists.

The call ties AI to both the creative pitch and the cost targets of the newly combined Paramount and Warner Bros. Skydance's Q3 investor materials carry the full shareholder letter and call.

Ogilvy and CAA Launch Echo to Turn One Celebrity Shoot Into 30 to 50 AI Ad Assets, With Terry Crews Opted In

Ogilvy and CAA launched Echo, a platform that builds AI ads from the face, voice and body scans stored in The CAA Vault, with Terry Crews among the talent who have opted in. Talent owns the likeness and signs off on every asset before it runs.

Ogilvy's announcement frames Echo as a way to turn a single hero shoot into localized, platform-native extension assets, through a partnership between WPP and CAA. Ogilvy staff pull the scans into WPP Open for storyboards, mock-ups and final creative.

204 cameras behind the scans. The Vault, built for CAA by Veritone, captures voice, body language, facial expressions and walk. Clients refresh their scans every three to five years, and the material is already used for de-aging and post-production fixes in films.
Eight to 10 assets become 30 to 50. A traditional shoot can expand into roughly 30 to 50 assets, which the companies say could cut costs by as much as 75%, according to The Hollywood Reporter.
Consent and watermarks built in. Automated compliance engines check output against contracts and brand guidelines, and AI ads carry a watermark marking them as AI-generated and approved by the star, Digiday reports.

Ogilvy expects to start with social assets and says it will still recommend a traditional shoot when a job demands it. Reel 360 looks at what Echo means for commercial production.

🦸 Prelulu. Generates real-time facial expressions and lip-sync for live, interruptible video calls with custom AI characters.

🎥 HiAPI. Unified, OpenAI-compatible API for image, video, music and text generation models, with hosted MCP access and pay-as-you-go billing.

🌄 masoc. Lightroom alternative for Apple devices, offering on-device photo and video editing, RAW support, film looks, and AI-assisted adjustments.

Browse the full tool directory

ElevenLabs' New Scribe Models, ComfyUI and Google Flow Add Speaker-Labeled Transcripts, ProRes Exports and Per-Second Pricing

ElevenLabs' Scribe v3, ComfyUI v0.44.0 and Gemini Omni 1.2 Flash in Google Flow each move AI output closer to the editing timeline, with what ElevenLabs says are more accurate transcripts, ProRes files straight from a graph and longer single takes.

ElevenLabs Scribe v3. ElevenLabs released Scribe v3 and Scribe v3 Live, speech-to-text models it says make 43% fewer errors, with speaker diarization and custom vocabularies of up to 10,000 terms. For editors, that could mean more reliable transcripts of noisy on-set dialogue and interviews when building string-outs.
ComfyUI v0.44.0. The core update saves video as ProRes with a choice of chroma subsampling for NLE-ready files, and adds local MiniMax H3 support with context windows for longer clips and a latent upscaler. H3 Max tools such as Recast, Extend, Lip Sync and Camera Control arrive as partner nodes.
Gemini Omni 1.2 Flash in Google Flow. Google AI Plus, Pro and Ultra subscribers can generate continuous shots up to 15 seconds, everyone else up to 10, and generations bill per second at 3 credits per second of 720p up to 10 seconds. Omni Extend steers a continuation with up to three reference images, and our breakdown covers the full credit table.

Legs of Steel Builds an Alaskan Glacier in the Hyperbowl LED Volume for Ski Film We Can't Not

In a behind-the-scenes video, Legs of Steel shows how it shot the final sequence of ski film We Can't Not in the Hyperbowl, a 500-square-meter LED volume outside Munich, walking the cast from a practical dining room onto a camera-tracked virtual glacier. It's a clear look at blending practical set pieces with an in-camera Unreal Engine backdrop.

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

🎙️ Blackmagic's first microphone, the Blackmagic Ambisonic Microphone, records 7th-order spatial audio over one USB-C cable straight into URSA Cine Immersive files, from $1,909.

🎥 Former Sundance CEO Joana Vicente and Jason Kliot's Lisbon-based accelerator Open Cities named its first 11 projects, backing indie filmmakers with mentors, funding and tool support from Google, Luma AI, Beeble, ZeroSpace, Wonder Studios and others.

🎨 Pokémon anime studio OLM built an AI coloring tool that learns a show's look from three hand-painted frames in about 90 seconds and cut painting workload by about 30% in tests.

📱 Meta is testing a Series feature that bundles Reels into episodes, with Pixie USA's 20-episode holiday microdrama Christmas at Maple Hollow Inn, which launches November 16 in select markets, among the first titles.

🏗️ Dark Slope is building a Vancouver LED stage at North Shore Studios from Pixomondo's shuttered stage hardware, giving producers on the lot a virtual production volume when it launches in December.

⚖️ Blade Runner 2049 producer Alcon settled with Warner Bros. Discovery over claims WBD helped Tesla make an AI image mimicking a still from the film for the Cybercab reveal, while its copyright suit against Tesla and Musk continues.