iQIYI's AI Film Studio: A Workflow Blueprint for Creators

What iQIYI actually built

On November 2, 2026, iQIYI announced what it calls China's first AI film studio, based at its Shanghai Creative Center. Production on the inaugural title, an internet feature film called Cop in Trouble, began on October 28. The film follows Jian Zhang, a low-ranking constable whose fabricated story about catching a notorious outlaw starts to unravel as his fame and fortune grow. No release date or budget has been announced yet, and the brief doesn't give us one, so we won't guess.

The studio itself is a modest 110 square meters, but it's built for a specific kind of capture: multi-camera shooting, synchronized high-resolution video and audio, digital continuity logging, and real-time footage review. That's a real physical set with real actors, not a rendering farm pretending to be one.

The twist is what happens right after the camera stops rolling. Footage syncs automatically to iQIYI's production platform, NadouPro, which gives filmmakers access to nearly 30 AI models, including Seedance 2.5 and Kling 4.0, plus hundreds of production-focused AI agents and skills. Those tools don't generate the actors. They transform the footage around the actors.

The core idea: lock the performance, then let AI handle the world

Liang Zhu, iQIYI's Vice President and Head of Intelligent Production, put it plainly: an actor's performance is the core of any film or series. The AI studio is built so directors can preserve that performance while adjusting sets, lighting, costumes, makeup, and props more flexibly after the fact.

This is the opposite of asking AI to generate a character from scratch in every shot. Instead, you capture a real performance once, with real timing, real blocking, and real emotional beats, and then use AI as a post-production layer that changes everything around that anchor. The performance is the fixed point. The environment is the variable.

iQIYI isn't new to this general approach. The studio builds on the company's Hengdian virtual studio, which it says delivered production-efficiency gains of more than 30 percent. The Shanghai studio is a continuation of that bet, not a one-off experiment.

Why this matters even if you're not making a feature film

You don't need a 110-square-meter studio with synchronized multi-camera rigs to borrow the underlying logic. The workflow iQIYI is formalizing at scale, shoot the performance for real, then use AI to handle set, wardrobe, lighting, and background, is something short-form creators can already approximate with a phone camera and existing AI video tools.

The practical reason this works better than full AI generation is continuity. If you're building a recurring character or a consistent host for a series, generating that character fresh in every clip invites drift: faces shift slightly, proportions change, mannerisms don't carry over. But if you record a real person (yourself, a creator, a voice actor) on camera once, you lock in the thing that's hardest to fake convincingly and hardest to keep consistent: a believable, continuous performance. AI then only has to handle what it's actually good at changing: the environment around that fixed anchor.

This also matches what iQIYI is explicitly selling the approach as: a way to open up production for emerging directors and resource-constrained teams, especially for bringing imaginative worlds to life across genres. Replace 'emerging directors' with 'small creator teams' and the pitch holds up just as well.

A scaled-down version of the pipeline

You can't replicate NadouPro's nearly 30 integrated AI models or iQIYI's continuity logging system, and the brief doesn't claim those tools are available outside iQIYI's pipeline. But the structure of the workflow translates to a much smaller setup.

• Shoot the real performance first. Get timing, blocking, and emotional delivery right on camera, even with a simple single-camera phone setup, before touching any AI tool.
• Treat that footage as the anchor. Don't regenerate the performer from scratch in post. Use the original footage as the reference point for everything downstream.
• Use AI for what's expensive or impractical to shoot for real: swapping backgrounds, adjusting lighting mood, changing costume or set dressing, or building environments you couldn't otherwise afford to build or travel to.
• Keep a simple continuity log. iQIYI uses digital continuity logging at studio scale; at creator scale, this can be as basic as a shot list noting which take, costume version, and background choice you used, so a recurring character stays visually consistent across episodes.
• Review footage in near real time. iQIYI's studio supports real-time review for a reason: catching a bad take before you've built AI post-production work on top of it saves time later.

What's still unproven, and what to watch

It's worth being honest about what's confirmed here and what isn't. iQIYI has announced the studio and started production on Cop in Trouble, but there's no finished film yet, no release date, and no budget disclosed. The 30 percent efficiency gain cited comes from the earlier Hengdian virtual studio, not from this new Shanghai facility specifically, so it's a reasonable signal of iQIYI's track record rather than a confirmed result for this project.

What's genuinely new is the packaging: a dedicated physical studio purpose-built to feed footage directly into an AI post-production pipeline, rather than AI and live-action being separate, loosely connected processes. For creators building a consistent cast or recurring character across episodes, this is a useful validation of an approach worth testing now. Platforms like Episodo that help creators manage a consistent cast across episodes are built around a similar premise: a stable, recognizable performance matters more than regenerating a character from nothing each time.

The thing to watch next is whether Cop in Trouble actually ships, and whether iQIYI shares any details about how much of the final film is AI-transformed footage versus raw capture. That will tell us how far this pipeline can be pushed before the live-action anchor stops being recognizable.