How Florida Universitària Cut 3D Previs Time by 90% with Meshy

A two-week pilot with 49 students at Florida Universitària in Spain cut placeholder modeling from 1–2 hours to 5–15 minutes, generating 448 models across 8 game development teams.

Pau Comes teaches 3D animation, games, and interactive environments at Florida Universitària, a higher education school in Valencia, Spain, where second-year students spend six months taking a game project from concept to publication. Inside that schedule, previsualization is the phase that suffers most.

To test whether generative AI could change that, Pau, together with another professor, ran a two-week pilot in November 2025, inside the Audiovisual Animation Project module, using Meshy 6 Preview to bring image-to-3D AI into the previs phase. With unlimited studio access for the duration, the students who adopted Meshy stopped building placeholder assets by hand. Instead, they fed their own concept art into Meshy, generated quick volumetric proxies, and used them to test cameras, scale, and composition in the actual scene, all before any final modeling work began.

That shift did more than save time. It pulled previs back into the curriculum as a real working phase and opened it up to the whole team, not just the one or two specialists who used to own it. It also gave the cohort their first hands-on experience with AI-assisted 3D pipelines, the same kind studios are now building.

Why 3D Previs Becomes a Bottleneck in Animation Classrooms

In animation and game education, pre-production is usually the first thing to go when deadlines tighten. Storyboards, animatics, and 3D previs sound essential on paper, but students who are still learning the full pipeline tend to jump straight into modeling final assets. Previs either shrinks into a rough sketch or drops out of the schedule altogether.

For Pau, that was the most fragile point in the curriculum. Before the pilot, his students kept running into the same four problems during early-stage 3D work:

• Slow manual placeholders. A comparable placeholder built by hand took 1 to 2 hours, leaving little time to test cameras, scale, and composition.
• Concentrated workload. With only a few hours budgeted for previs, one or two students per group would produce a single version, while the rest of the team stayed locked into their own specializations.
• Hard-to-communicate spatial ideas. Sketches and verbal descriptions weren't enough to get a team aligned on form, scale, and framing before production.
• Imagination ahead of skill. Some students could clearly picture a scene but couldn't yet model it from scratch, so good ideas stayed in their heads.

The result was a familiar gap between storyboard and blocking. That gap is exactly where better image-to-3D tooling makes the most difference.

Designing a Two-Week Pilot Inside a Real Game Project

Pau ran the study inside an active production context, not as a standalone exercise:

• 49 second-year students in the Higher Degree Vocational Training in 3D Animation & Interactive Environments (Spain's Formación Profesional / Vocational Training system)
• 8 development teams, each producing a cinematic or game trailer for their own title
• A two-week previs window inside the wider six-month project
• Unlimited Meshy access in studio mode for the duration
• Mixed-methods evaluation: pre- and post-pilot surveys, in-platform asset logs, and qualitative review of generated models and team outputs

The scope was clear from day one. Meshy was scoped specifically to the previs phase from day one, where speed and iteration matter more than final production quality. The question was narrower: could AI keep previs from being sacrificed, and help students make better decisions before committing to hundreds of hours of modeling?

How Students Used Meshy's Image-to-3D for Previs

A clear pattern emerged within the first few days. Rather than working with text prompts alone, students used Meshy as a 2D-to-3D translator: they fed in their own concept art and earlier sketches, then evaluated whether the volumetric interpretation that came back was worth taking further.

The image-to-3D workflow most teams settled on:

1. Upload reference. Concept art or a 2D sketch from earlier in the project.
2. Generate a volume. Image-to-3D returns a quick textured proxy in minutes.
3. Drop it into the scene. Test proportions, framing, and scale against the rest of the layout.
4. Decide. Keep what works, clean up topology where needed, and move into production with a clearer sense of how the final shot should look.

What Meshy gave them was a volumetric sketchbook: the 3D snapshot a previs phase actually needs. Teams could try a framing, throw it out, try another, and reach the production phase with decisions already locked in instead of guessed at.

A few use cases came up across almost every team:

• Populating scenes with textured proxies to move beyond a flat greybox
• Testing camera angles and scale against actual volumes instead of imagined ones
• Converting 2D concept art into 3D mock-ups the whole team could critique
• Communicating spatial ideas that were hard to describe verbally or sketch quickly

Throughout, the student stayed in charge of the work. Meshy proposed, and students selected, corrected, and decided. That mix of fast AI generation with consistent human judgment was what made the workflow stick.

"Meshy helped students clarify ideas and show them in an image."

Pau Comes

3D Animation Professor, Florida Universitària

What Meshy Changed in the Classroom

Evaluation at a Glance

After the two-week trial, students rated Meshy across six dimensions:

The scores landed exactly where the pilot was designed to test. Ease of use, utility for previsualization, and speed — the three dimensions that matter most for moving from concept to a working 3D layout — all rated above 4 out of 5.

Faster Previs Restructured How Teams Worked

The biggest gain was speed. Placeholder creation dropped from 1–2 hours to roughly 5–15 minutes per model, a reduction of up to 90%. Across the two weeks, the eight teams generated 448 models between them.

The more interesting change was structural. Faster placeholders let teams rework how they divided up the early stage:

• Sub-teams produced parallel previs versions and compared them side by side, instead of committing to a single rushed version up front.
• **Every student could contribute to early visual decisions,**not just the one or two specialists who used to own previs by default — though in practice, usage concentrated among a few heavy users.
• More project time went into camera and rhythm experiments, with more instructor feedback per iteration.

For a cohort built around cross-disciplinary teamwork, that last point mattered as much as the speed itself. Students were no longer pushed into narrow roles tied only to their specialization, and ownership of early creative decisions ended up shared across the team.

A New Role: From Modeler to 3D Director

The pilot also changed what students were doing during previs. Less time went into modeling from scratch at this stage. More went into framing prompts, choosing between variations, critiquing outputs, and planning corrections.

That shift opened up room to teach skills the industry is already asking for:

• Multimodal prompt design. Working fluently with both text and image inputs.
• Curating AI outputs. Knowing when a generation is useful as a reference and when it isn't.
• Retopology and cleanup of AI-generated meshes. A concrete technical skill layered on top of the generative workflow.

The role moved partway from modeler toward 3D director: someone who frames the problem, evaluates the options, and directs the refinement.

AI as Cognitive Scaffolding: Accessibility in 3D Education

One finding reframed how the team thought about adoption itself. A learner who typically struggled with manual 3D modeling generated 57% of all models in the pilot. For some learners, generative AI works as a kind of cognitive scaffolding: a bridge between imagination and execution that lets them externalize ideas they can't yet model by hand.

That points to a real accessibility benefit alongside the productivity one. AI 3D tools can widen who gets to contribute to early creative work, not just speed up the people already producing it. For a program built around cross-disciplinary teamwork, that's a meaningful shift. Students who might otherwise have handed modeling tasks off to teammates were able to take part directly in shaping the visual direction of their project.

"It helps me express what I don't know how to model, or saves me hours of searching." one student participant said.

A Realistic Look at Adoption

Adoption inside the cohort was uneven, and that turned out to be useful. About 55% of students embraced Meshy, while 45% chose not to use it, mostly for considered reasons rather than technical barriers: a preference for manual craft, concerns about AI's impact on creative work, or questions around copyright and style. The teaching team built that split into the module itself, treating it as a chance to discuss how to coexist with AI without losing artistic identity. By the end of the pilot, attitudes had moved from broad anxiety to a clearer sense of where the tool actually fits.

What's Next: Building Meshy Into the Program

The pilot only covered pre-production. The next phase, already underway, takes the previs assets generated with Meshy and uses them to lock in camera, scale, and narrative rhythm before producing final assets through the traditional pipeline. The difference, compared to previous cohorts, is far less uncertainty going into production.

For Florida Universitària, the pilot has led to several concrete decisions about the program itself:

• Position Meshy as a concept and 3D previs tool inside the curriculum: a rapid mock-up and volumetric sketching layer, with room to evolve into final production over time.
• Design "Generate + Fix" exercises where students generate a base in Meshy, retopologize and UV-unwrap in Blender, and integrate the result into Unity or Unreal.
• Add AI literacy and ethics modules to the program, so students learn to use these tools critically rather than uncritically.
• Extend the approach to other subjects. Digitization courses and more advanced programs are under consideration, possibly as a workshop or bootcamp format.

For a teacher working with tight schedules and demanding students, Meshy found its place exactly where the curriculum was most fragile: previsualization.

"The question isn't whether AI will enter the classroom. It's what we want it there for."

Pau Comes

3D Animation Professor, Florida Universitària

In this pilot, the answer was clear: a catalyst for the most fragile phase of the creative process.

Florida Universitària is now running the next phase of the experiment with this year's cohort. We'll be back with what they find.