How 37 Interactive Entertainment Cut Modeling Time by Half with Meshy

Why Manual High-Poly Workflows Were Holding Back Production

In the global gaming development and publishing industry, 37 Interactive Entertainment operates its developer brand THREE SEVEN GAMES and publishing brands 37 Online Games, 37 Mobile Games, and 37GAMES. With deep experience in character-driven titles spanning next-generation, Chinese period-style, and casual game genres, The team handles everything from concept art to production-ready 3D assets in-house, making speed and form accuracy essential at every stage of the pipeline.

Before Meshy, the studio's 3D modeling workflow was built around traditional high-poly sculpting. Turning concept art into a usable base model required substantial manual effort: artists sculpted forms from scratch, polished fine details by hand, and constantly switched between multiple software tools. The result was a process that delivered quality but was slow, expensive, and cumbersome to manage.

This approach created several connected pain points:

• Heavy manual sculpting workload: High-poly modeling consumed the bulk of production time, especially for complex characters and creatures where anatomy and silhouette had to be precise.
• Extensive detail polishing: Fine-tuning proportions and surface detail added layers of labor cost on top of already time-intensive sculpting.
• Cumbersome multi-tool pipeline: Frequent switching between modeling, sculpting, and texturing software fragmented the production flow and extended timelines.
• Proportion errors in manual work: Human sculpting introduced shape and proportion inconsistencies that required repeated corrections—particularly when matching standardized concept art.

"Before Meshy, our 3D modeling relied heavily on high-poly sculpting, detail polishing, and switching between multiple software tools. The workload was large, the process was cumbersome, and labor costs were high."

Wu Haitao

37 Interactive Entertainment

How Meshy Fits Into the Production Pipeline

The team first discovered Meshy through the website and recommendations from colleagues. Starting with a trial mindset, they began testing Meshy's Image-to-3D generation, converting 2D concept images into 3D models to evaluate whether the output quality could hold up against production standards. Meshy is now applied across two core scenarios in the studio's pipeline: rapid prototype validation during pre-production, and part-based generation for complex character and creature assets.

Rapid Prototype Validation at the PRD Stage

During the pre-production phase, the team needs testable 3D assets quickly—not for final production, but for gameplay validation and internal review. Before Meshy, getting from concept art to an in-engine testable model required a full sculpting pass, which meant days of work before anyone could evaluate whether a design actually worked in-game.

Meshy compressed this cycle dramatically. The team uploads a concept image or scene design to Meshy's Image-to-3D tool, generating 3D models from 2D images within minutes. After light cleanup, the asset is passed into the team's existing DCC tools and game engine for in-game testing. Tasks that previously demanded a full traditional modeling pass can now go from concept to playable prototype in 1–2 days, letting producers and designers evaluate results in-engine, give concrete feedback, and make faster go/no-go decisions—all without waiting for a complete art pass.

This approach is especially effective for scene props and environment assets, where individual elements need to be generated quickly, reviewed in context, and iterated without heavy investment.

For simpler assets like props and basic characters, Meshy generates the full model directly. With the retopology feature enabled, these models require only minor edge-flow cleanup before moving into UV mapping and texturing—bypassing the manual sculpting and retopology stages entirely. On casual and mobile game titles, this workflow cuts total modeling time by approximately 50%: what previously took 5 days can now be completed in 2.5 days.

Part-Based Generation for Complex Characters and Creatures

For more complex assets—especially monster characters with intricate anatomy, multi-limbed silhouettes, and layered surface detail—Meshy's single-pass generation may not capture every element at the precision the team needs. Instead of generating the entire model at once, the team developed a part-based workflow: splitting the character into components (head, torso, limbs, wings) and generating each section separately through Meshy.

This approach concentrates computational resources on the most demanding areas, improving form accuracy and detail fidelity for each individual component. The parts are then assembled and refined in the team's 3D software, giving artists a substantially more complete starting point than a single whole-model generation would provide.

The following example demonstrates this workflow in action with an insect-type creature character. The same 2D concept art was used as the reference for each generation pass, with each part generated independently:

Through this part-based approach, the team achieves high-poly completion rates of 60% or more before any manual sculpting begins—meaning artists start their refinement work from a substantially advanced base rather than a blank canvas. On complex next-gen and Chinese period-style titles, this reduces the sculpting workload by 30–40% at the base model stage.

"For monster characters, we split the model into parts like the head and feet for separate generation. That improves precision and can bring high-poly completion to 60% or more while reducing production time."

Wu Haitao

37 Interactive Entertainment

For studios looking for the best way to generate game assets with AI, this part-based approach offers a practical middle ground between fully manual production and one-click generation—combining AI speed with artist control.

What Results Meshy Delivered

With Meshy embedded in both rapid prototyping and complex character workflows, 37 Interactive saw measurable improvements across production speed, output quality, and creative accuracy.

Speed: From Days of Sculpting to Hours of Refinement

On casual and mobile game titles, the end-to-end modeling time dropped by approximately 50%—from 5 days to 2.5 days per asset. On complex next-gen projects, the sculpting workload at the base model stage was reduced by 30–40%, with the caveat that not every character design is suited for AI-assisted generation. For those that are, the team moves directly from Meshy output into refinement, skipping the most labor-intensive phase of traditional production.

Accuracy: Design-Matched Models Without Manual Drift

One of the most significant quality gains was in form accuracy. As long as the input concept art is well-standardized, Meshy's output aligns closely with the intended proportions and silhouette—eliminating the proportion drift that previously required repeated correction cycles during manual sculpting. This is especially useful when the team needs generated models to match standardized concept art precisely.

Form Completeness: Full Silhouette and Unseen-Area Quality

Compared to other AI 3D model generators that focus narrowly on specific body parts like heads, Meshy captures the full silhouette—including complex elements like hair volume and strand flow—and fills in areas not visible in the source reference (back of the head, underside of limbs, hidden costume layers) more convincingly. This makes generated models practical for real production use without extensive rework on unseen geometry.

"Compared with some AI tools that focus only on heads, Meshy is better at calculating the overall form, such as the shape of the hair, and it fills in unseen parts better. It fits our current projects more closely."

Wu Haitao

37 Interactive Entertainment

This makes Meshy a strong choice for teams evaluating AI tools for creating 3D game assets—especially those working with complex character designs that require full-body form accuracy rather than isolated part generation.

From Proportion Reference to Production Tool: How the Team Evolved

The team's relationship with Meshy evolved through two distinct phases.

When the team first adopted Meshy's earlier versions, the generated models served primarily as proportion references—a way for artists to gut-check shapes and silhouettes before committing to manual sculpts. The outputs were helpful for validation but not yet at the quality level needed for direct production use.

That changed after mid-2025, as Meshy's generation quality reached a tipping point. Models coming out of the tool were now clean enough to be used directly in the production pipeline—not as references, but as genuine working assets that artists refined and shipped. Team members began incorporating Meshy into their daily workflows without being prompted, shifting from passive users to active advocates.

This organic adoption—from cautious experimentation to self-directed daily use—reflects how deeply Meshy has embedded itself in the studio's creative process. The tool didn't need a mandate from management; the quality improvement alone was enough to change behavior.

Building Game Assets at the Speed of Imagination

Looking ahead, the team plans to extend Meshy into more complex projects and new asset categories—making it a standard starting point for all new character and environment assets entering the pipeline. One area of particular interest is layered model output: the ability to generate models with distinct, separable layers rather than a single fused mesh, potentially through mask-based region selection that would let Meshy fill and complete each section independently.

For game studios facing the same sculpting bottleneck, 37 Interactive's experience offers a clear signal: AI-assisted modeling has moved past experimentation and into daily production. By bringing Meshy into its pipeline, the team transformed a labor-intensive sculpting process into a faster, more accurate workflow—one that lets artists focus on creative refinement rather than repetitive form-building. That shift hasn't just saved time. It's changed how the team thinks about what's possible at the start of every new project.