Academy · 3D tools report

Six 3D tools, tested

For characters and for scenes: what each tool is good at, where it falls short, and what to use next.

Characters

Tripo Studio · Tripo Canvas · Charakuru · VRoid

Scenes

World Labs Atlas · Pure code (three.js) + Tripo Studio props (direction, testing)

At a glance

Six tools side by side

Tripo Studiohead-swap workflowTripo CanvasCharakuruVRoidWorld Labs AtlasPure code
Used forCharacters
+ scene props, testing
Characters, propsAnime charactersAnime headsScene previewsScenes
with Tripo Studio props
OutputRigged GLBSame as StudioVRMVRMGaussian splatthree.js scene
Usable in the game✓✓✓✓✗
look only
✓
Claude can run it✓△
agent misreads
✓✗
hand-made in UI
△
frame by frame
✓
Cost~440 credits
per character
Same prices
images free
A few hundred creditsFree~$7
per room
No credits
VerdictRecommendedSame as StudioCandidateBackupPreviews onlyStill testing
scene direction

Character tools are tested end to end. Scene tools are not: the current scene direction is Tripo Studio + pure code, many scene tests are still to run, and the scene workflow will differ from the character one.

Tripo Studio · characters

The head-swap workflow Recommended

Head, hair and body are modelled separately, then put together.

1
Generate four imagesFull body, bald full body, bald head, hair front and backGPT image
2
Model head and hairSmart Mesh, quads → Smart UV → de-lit textureTripo
3
Fit hair to headMeasured against the source imageClaude + Blender
4
Model, rig and animate the bodyVRM humanoid skeleton, six preset animationsTripo
5
Swap the headOn the rigged file; skeleton and animations untouchedClaude + Blender
6
Fix the animationsDrooping shoulders, toes turned out, clippingClaude + Blender
Tripo generates. Claude and Blender assemble and fix.

Step by step: academy-character-workflow.pages.dev

Tripo Studio · characters

Two characters, two art styles

Avatar Girl: round avatar reference, AI-generated full-body image, finished 3D model Belle: trailer frame reference, AI-generated full-body image, finished 3D model
~440 creditsTripo, per character (estimate)
6 animationseach, picked from 97 free presets

Both characters side by side: academy-character-preview.pages.dev

Tripo Studio · characters

Strengths and gaps

✓ Strength✗ Gap
Art stylesWorks for both a 3D-render look and an anime face
ControlEach part on its own: the face stays sharp, the hair can get physics. A person reviews every step, so quality stays under controlMany manual steps; labour-intensive
Animation97 preset animations, free to applyThe presets share the same faults (drooping shoulders, toes out), fixed every time
ReuseEvery fix is a script, ready for the next character
BudgetThe two characters are 120,000–130,000 triangles; the 25,000 budget is tested on one head only
FinishingNo VRM packaging, expressions or toon shading yet
LicenceNo issues
Key lesson. Measure every hand-tuned value against the image that was fed to Tripo, before anyone else looks at it.
Tripo Studio vs Charakuru

Same trunk, different ends

Both go image → Tripo Smart Mesh → assemble → rig → animate. Charakuru adds a VRM finishing stage; the head-swap workflow does more in the front half.

CharakuruHead-swap workflow
PartsHead with hair, body · 2Bald head, hair, full body · 3
Candidates4 per part, pick one1 per part
Polycount25,000 quads per partPer part: head 2,500 · hair ~4,500 · body ~5,000 ✓
AnimationsOne bundled walk, not usable outside the kit97 Tripo presets ✓
Hair physicsAutomatic ✓Added with scripts
Toon shadingBuilt in ✓Not yet
ExpressionsBuilt in ✓Not yet
OutputVRM file ✓GLB with VRM bone names
LicenceWritten licence for commercial use ✗No issue ✓
Keep the head-swap trunk, borrow Charakuru's last stage (VRM, expressions, toon shading) with in-house scripts.
Tripo Canvas

Better visibility, not better models Same as Studio

Tripo CanvasClaude driving Tripo Studio
Models and pricesSameSame
Image generationBuilt in, 0 credits on the Max plan ✓Codex or OpenAI API, ~$0.10 per image
Visible to the teamNode graph, saved as a template ✓In chat and scripts
Following instructionsThe agent often misreads ✗Stable ✓
Checks against the sourceNone ✗Proportions, overlays, clipping, every frame ✓
PitfallsEnter sends; a second upload doesn't attach; no price on confirm cards
Claude's workflow is more stable. Later, turn the working pipeline into a Canvas template for the team.
Tripo Canvas · AI 3D Lab

AI 3D Lab vs Claude + Blender

The same kind of thing: an AI agent that operates 3D software from written instructions. Claude + Blender is more controllable.

AI 3D LabClaude + Blender
Where it runsInside Tripo, no export neededLocally, after export from Tripo
What it didNot visible ✗Every step is a script ✓
ReuseNew instructions each timeScripts reused for the next character ✓
CheckingScores its own similarity to the referenceMeasured: proportions, overlays, clipping, every frame ✓
Knows the projectOnly this task's inputsBudgets, lighting rules, VRM skeleton ✓
CostCredits per task, price not publishedNo Tripo credits
Charakuru

Anime illustration to VRM, in one kit Candidate

Base Body C made with Charakuru: turnaround and seven expressions

Strengths

  • Hair and skirt physics, toon shading and expressions built in
  • Scriptable end to end; cheap

Weaknesses

  • Commercial use needs a written licence for companies with ¥100M+ in annual revenue or total funding
  • No 3D refinement stage
  • The body has to be split into upper and lower halves
Anime characters△ Worth a retest with an anime character
Non-anime characters✗ Use the head-swap workflow instead
Base Body C (made)✓ Usable

The test used a bald, non-anime base body, so its strengths went unused.

VRoid

The textbook anime face Backup

VRoid base model with anime shading: front, three-quarter, side, and hard toon shading with side light

Strengths

  • Flat face with toon light: clean, anime-style shading
  • Native VRM; the base model is free

Weaknesses

  • Edited by hand in VRoid Studio's interface only
  • Every character starts from the same base face
  • Complex outfits are out of reach
Fallback for anime heads✓ If Smart Mesh falls short
Main workflow✗ Tripo Smart Mesh already makes anime heads (Belle)
Its toon shading✓ Worth borrowing
World Labs Atlas · scenes

Closest to the concept art, but look-only Previews only

Concept image, front half of the room
Input · concept 1
Concept image, back half of the room
Input · concept 2
→
Atlas splat, entrance looking in
Output · a new angle
Atlas splat, fireplace end looking back
Output · a new angle
Atlas
Fidelity to concept✓ The highest of all
Fills in the unseen✓ A whole room from two images
CostAbout $7 per room
What it is✗ A Gaussian splat: nothing moves or can be reused
Into 3D models✗ A big job, close to remodelling
Weight and effort32 MB per room; every new view placed by hand
Use it for
Promo shots✓
Space previews for the team✓
Game levels✗ No movement, no collision
Every room this way✗ Too heavy, too slow
Pure code · scenes

A walkable room and exterior Still testing

Concept image
Concept
Room written by Codex, reviewed by Claude
Codex writes, Claude reviews
Room written by Claude
Claude writes
Clubhouse exterior written by Claude
Exterior · Claude
Codex writes, Claude reviewsClaude writes
Scale25 furniture types46 object types, 104 placed
Download2.5 MB2.6 MB
Review score6 / 107 / 10 room · 6 / 10 exterior
Heavy partPer-pixel shadingTextures in GPU memory
Known so far
Collision, interaction✓
Plants and fabric✗ The weakest part
Indoor ↔ outdoor, door transition△ Still testing
Mixing in Tripo Studio props△ Still testing
Phones△ Still testing
Scenes · status

Scenes are still early Many tests not run yet

The scene conclusions here are provisional. The scene workflow will not look like the character workflow: a room needs a shell, exact sizes and both sides of every wall, and image-to-3D does not produce that. Tripo's own interior tutorial builds the room shell with code or AI 3D Lab, not from an image.

Current direction: Tripo Studio + pure code

Not settled yet. The direction being tested next.

Pure code

The room shell and the building: floors, walls, roof, doors, windows, plain furniture

Tripo Studio

Props whose shape matters: plants, statues, instruments, robots

TestWhat it answersStatus
Tripo Studio Smart Mesh for propsHow good and how heavy AI-made props are inside a room△ Parts list only, no models yet
Tripo Studio props inside a code-built roomWhether style, scale and lighting match✗ Not started
Indoor ↔ outdoor, door transitionHow a player gets from the exterior into the room△ Not decided
Phones and larger areasFrame rate, memory, loading as the player moves✗ Not started
Expect a separate scene workflow once these tests are done. Only one thing is settled: Atlas for promo shots and previews.
Recommendation

What to use next

UseStatus
CharactersThe Tripo head-swap workflow; add the last stage (VRM, expressions, toon shading) the way Charakuru does itSettled
Anime charactersCharakuru as a candidate, after a retest; VRoid as a backupTo retest
Scene previewsAtlas, for promo shots and space previews onlySettled
Scenes in the gameDirection: Tripo Studio + pure code (code for the room and building, Tripo Studio for props). A separate scene workflow will follow the testsStill testing
Division of workTripo generates; Claude and Blender assemble and fix, measuring against the source imageSettled
Next

Next steps

#TaskWhy
1Build one character end to end within budgetCheck the polycount: head 2,500 · hair ~4,500 · body ~5,000 quads
2Add VRM, expressions and toon shadingThe last stage of the character workflow
3Retest Charakuru with an anime characterThe first test used the wrong kind of input; ask about the licence at the same time
4Make a Canvas templateSo the team can follow the workflow
5Test the Tripo Studio + pure code directionTripo Studio props in a code-built room, the indoor–outdoor connection, phones; then write the scene workflow