Control beyond text prompts
Expect more interfaces for directing generative output: virtual camera rigs, joysticks, motion tracks and camera-parameter uploads. This is already visible in some tools that accept camera movement or motion data to guide renders. The hosts want real-time control and richer input modalities beyond typed prompts.
Photoreal image models without uncanny valley
One confident call: image models will reach consistent photorealism, producing people and scenes that are visually indistinguishable from real photography. The hosts argue we’re close now and that 2026 could be the year this becomes routine.
Very likely — shifts that will influence strategy and budgets
Resurgence of handcrafted craft
As AI-generated material floods feeds, the hosts predict a counter-movement: renewed interest in tangible, handcrafted media — stop-motion, hand-drawn 2D, film formats, collector editions and physical releases. This is a cultural and market reaction, not a technological one.
Nvidia and a market recalibration
With massive valuations tied to AI compute, the hosts predict a market correction that reflects the reality that AGI is still years away. This could affect investment flows into chip suppliers and cloud compute providers.
Legal clarity on model training
Ongoing litigation around what data models can be trained on looks set to move toward clearer rulings or settlements. The discussion will shift from training legality to how outputs are used and whether they infringe IP.
A non-US model that runs locally
Expect a surprise entrant from outside the U.S. that offers high-quality image and video models optimized for local execution, compact and Comfy-friendly. This would echo past moments when non-US models disrupted assumptions about who controls core tooling.
AI-assisted animated features arrive
The hosts call an AI-heavy animated feature likely within the next year — one that leans on AI for inbetween frames, style passes, and massive cost-time reduction, and that could appear at a major festival or streamer.
New paradigm for persistent AI-generated worlds
Following demos of world models, expect usable, persistent generative spaces where a virtual camera can move and capture shots. Current demos are short-lived; the shift will be toward duration, state persistence and higher fidelity.
Long shots — ambitious changes worth watching
90-minute live-actiony cinema from AI studios
Several AI-first studios are attempting full-length projects. The hosts label a true live-action feature driven by generative tech as a long shot for 2026: possible, but still constrained by uncanny valley and performance synthesis limits.
Neural rendering integrated into major DCC tools
The idea: replace or augment ray-traced rendering with neural render passes inside Maya, Blender, Houdini or Unreal. This is a technical and product effort that could land if compute and quality economics improve.
Licensed AI music ecosystems
Partnerships between music licensors and AI platforms could mainstream licensed AI music. If labels cooperate, training datasets become licensed and “ethical” offerings expand. Expect ambient and background music to be largely AI-generated, while marquee artists remain human-driven.
Closing thoughts
The hosts distilled a practical, production-first view of the year ahead: incremental, impactful tool releases will change how shots are planned, how VFX is executed and how animation budgets are allocated. Major structural shifts — AGI, full neural replacement of render engines, and studio-level live-action features driven end-to-end by generative models — remain less likely in 12 months but are plausible longer-term bets.
The next milestone is simple: test the new tools in small, production-representative pilots and update pipeline docs as capabilities and vendor terms evolve. 2026 will be about integrating practical AI into established production rhythms, not about overnight reinvention.
