Animated Company founder and CEO Douglas McGinness III examines how AI tools are reshaping traditional visual effects workflows, from concept design and previs through roto, matchmove, and final delivery, while raising critical questions about authorship, copyright, and the preservation of storytelling craft.
VFX has always been constrained by human bandwidth and sequential workflows.Highly specialized artists, complex tool chains, and global teams meant that even simple changes cascaded through roto, layout, lighting, comp, and delivery.
The work is beautiful, but the process is slow and expensive because so much of it has historically been manual, pixel-level, and difficult to parallelize.But AI is changing all that.It’s not some magic solution, but rather a wrench that’s been thrown into a century-old gearbox, making things work faster, cheaper, and more creatively, heralding the birth of the bijou design studio.
In the next five years, the production line approach will be gone, replaced by small, iterative teams making studio-level work.The coming impact on the industry The Achilles heel of VFX has always been speed.There are multiple reasons for this, but primarily because only a small set of expensive, talented artists and technologists could do the work. Key tasks, like roto, paint, matchmove, and plate prep have also historically been heavily manual.
And pipelines are fragile: one change ripples downstream into re-renders, re-comps, and additional quality control.Major entertainment studios are now running full-pipeline AI R&D projects – from concept design through production and post – to see where AI can both cut costs and expand output by compressing schedules.Inside those experiments, you see three big patterns: First, we’ll see a change in concept and design.
Image models are increasingly used alongside or instead of pure Photoshop workflows for key art, environments, and character exploration.Artists can generate dozens of variations on a direction in hours instead of days, then paint over and refine the strongest options rather than starting every frame from scratch.Second, this will lead to the move from storyboards to animatics.
Off-the-shelf video models, such as Google’s Veo 3, can now turn boards and style frames into moving animatics with sound, allowing directors and clients to see pacing, framing, and tone long before traditional previs or animation is commissioned.Veo 3 produces 4K, physically coherent motion and synced audio, and is already integrated into tools like Gemini and YouTube Shorts for fast visual iteration.Lastly, all of that opens the door to end-to-end experiments.
Campaigns like Coca-Cola’s 2025 “Holidays Are Coming” remake, produced with AI studio Secret Level, show what a fully generative pipeline looks like at global brand scale: tens of thousands of AI-generated clips iterated by a relatively small core team to produce an all-CG Christmas spot.The ad cut production time and allowed massive variation, but it also triggered backlash from audiences who felt the result lacked the warmth and charm of earlier live-action versions… which raises the following question: Will lower costs mean lower production values? That’s the fear, isn’t it? It’s the fear with all AI and tech-based solutions.You save time and lower costs, but you pay for it in other ways.
While the Coca-Cola ad is an impressive demonstration of what a generative pipeline can do for a global brand, it lacked the emotional continuity that the public expects.So, the lesson maybe isn’t that AI lowers production values, but that production value is about more than the quality of the animation.In VFX, if you design the pipeline correctly, automating predictable, repetitive tasks and using the saved time to concentrate on story beats, creative direction, and design cohesion, the production values stay the same, because you’re putting your time into the things that viewers notice. The trick is to front-load creative decisions, industrialize only what survives that creative gate, and keep humans in the loop for final intent.
In other words, you need to leave people to handle the story, performance, composition, and final intent, and AI can take care of the more predictable elements.Can AI do the heavy lifting? AI is never a hands-off solution.You always need a human to write the prompts, guide the outputs, define intent, and curate inputs.
That being said, it can still do a lot.Roto and mattes – training show-specific models that deliver production-grade mattes from a limited hand-rotoscoped set Cleanup and inpaint – removing rigs, booms, or background clutter and reconstructing plausible plates for artists to refine Matchmove and plate intelligence – depth maps, camera path estimation, and relighting cues that augment tracking and layout Previs and animatics – turning scripts, boards, or stills into moving shots with Veo 3 and similar models, sometimes good enough to stand in for early marketing or internal pitches Stylization and look development – style transfer and fine-tuned models that enforce a show-specific look across many shots, with artists providing keyframes and corrections.So, you can model-accelerated pipelines with skilled artists doing direction, refinement, and QC, and producers who define clear sign-off points for AI output vs.
final frames.There are masses of potential for effective AI deployment, you just need to initiate processes to both maximize that potential and provide legal protection.Training foundation models for styles or IP When training foundation models for specific styles or intellectual property, you need to work with a number of key principles.
Clean-room training and provenance come first.Models should be trained and stored in controlled environments, with every step of the process carefully documented.That means maintaining a clear record of each dataset used and of any instance where consent was not obtained, along with the reasoning and mitigation steps.
The goal is to be able to demonstrate provenance, showing exactly what went into the model, in order to minimize any legal or reputational risks further down the road.Licenses and ownership are equally important.Any training corpus that includes third-party intellectual property should be supported by express licenses.
This removes ambiguity and ensures that all contributors understand the scope of use.Before training begins, decide who owns the resulting model – the client, the vendor, or both – so you can be certain that there will be no disputes once the system begins to generate value.And then we come to strong enterprise controls.
These are essential, especially when you’re working with recognizable names, brands, or production-ready material.By relying on enterprise-grade models and private deployments you can maintain confidentiality and control over data.Netflix provides a useful benchmark: disclose any planned use of AI early, avoid tools that automatically store production inputs, and never modify the likeness or performance of talent without explicit consent.
IP and copyright There is an ever-growing grey area when it comes to AI, IP, and copyright.Who owns what AI has produced? But although the legal landscape is messy, it’s still actionable; platforms and studios just need contractual guardrails to protect them.In reality, that means model training rights, provenance records, watermarking, and explicit clauses about character/talent replication, while practical safeguards include dataset curation, consent registries, signed approvals for any final-frame AI work, and robust audit trails.
The thing to remember is that IP becomes more valuable when “anyone can make anything with AI.” That’s why it’s so important to protect the story, the craft, and permission to do it.AI may be democratizing animation – Sora 2 can create near-perfect animation with little human input.But with more people using AI to generate reproductions of established IP, including , Studio Ghibli, Nintendo, and , more needs to be done to prevent this theft.
Some of that comes down to the introduction of both platform and industry guardrails, detecting and blocking clear IP misappropriation.But there’s also more that businesses can do themselves, through the creation of IP-safe models and services, whether that’s licensed style packs, bespoke fine-tuned models sold/licensed by rights holders, or clean-room production options for clients who need legal guarantees.That AI should change VFX is inevitable.
And there will always be those who say it’s a bad thing.But used properly, AI is and will be transformative.It’s not about lowering production values, just the reallocation of human effort, replacing grunt work with higher-level creative and narrative decisions.
--
Douglas McGinness III is founder and CEO of Animated Company, a creative technology studio and toolmaker that creates premium animated worlds for brands and original IP using proprietary workflows.Their Fossa AI animation tool suite solves specific bottlenecks in animation and motion graphics design.Douglas McGinness III is founder and CEO of Animated Company, a creative technology studio and toolmaker.
Masterpiece X Launches WorldEngen AI Copilot
Emmy Winning Pixar Vet Stephan Bugaj Releases ‘The Seeker’ AI-Generated Film
Adobe, Runway Partner For Next Generation of AI Video
Adobe Firefly Introduces New AI Video Tools, Partner Models
Chaos Delivers Maya, Houdini, Real-Time Rendering and AI Tools to Blender
Wētā FX, AMD Partner to Develop Next-Generation VFX Tools
Wētā FX and AWS to Develop AI Tools for VFX Artists
NVIDIA Studio Presents ‘Hologram’ Crowdsourced Music Video
