Meshy 3D Agent's rig wizard dropped 13 markers on my coffee-brand robot and got both wrists wrong: one sat on the shoulder, the other floated outside the body. Its Symmetry toggle, on by default, turned my correction of one wrist into a worse spot for the other. Message 3 alone cost me about 15 clicks and drags on Meshy. Snapshapy's 3D Agent rigged the same request with zero markers and zero clicks.
I tested both agents on my own paid plans: Meshy Pro ($20 a month, 1,000 credits) at meshy.ai and a paid Snapshapy account at snapshapy.com. I ran Meshy on 2026-11-03. I ran Snapshapy on 2026-11-04, a day later, after Snapshapy shipped an update. Both got the same three messages, typed verbatim: "Give me 4 concept options for a friendly robot mascot for a coffee brand.", "Go with option 1 and turn it into a textured 3D model." and "Rig it and make it wave and dance." On every choice card I picked the first option.
This review follows the third message, marker by marker. Two asides for the record: Meshy drew 1 concept of the 4 I asked for, and the full run cost 39 credits on Meshy against 14 on Snapshapy.
My verdict in 60 seconds — Meshy 3D Agent 57/100, Snapshapy 88/100
Meshy 3D Agent does get a coffee robot to wave, but it hands the hardest part of rigging back to the user, and two of its defaults made that part harder.
- Rigging its own model — Meshy 2/10 · Snapshapy 9/10 — refused at about 903,000 faces vs a 300,000 limit; Snapshapy reduced to 100k without asking
- Who picks the remesh — Meshy 3/10 · Snapshapy 10/10 — a remesh card plus a "Rig Model" confirm, 3 clicks vs 0
- Speed — Meshy 8/10 · Snapshapy 5/10 — 92 s vs 291.7 s to 3D, ≈305 s vs ≈475 s in total
- Default marker placement — Meshy 2/10 · Snapshapy 10/10 — 2 of 13 markers wrong at the wrists vs 0 markers
- Symmetry default — Meshy 3/10 · Snapshapy 10/10 — 1 correction mirrored into a worse spot, 1 extra click to switch it off
- Clicks after "rig it" — Meshy 2/10 · Snapshapy 10/10 — ≈15 vs 0 for the same sentence
- Motions per file — Meshy 3/10 · Snapshapy 9/10 — 1 per file in 2 files vs 5 clips in one 47.4 MB file
- Ten mascot variants — Meshy 2/10 · Snapshapy 10/10 — 10 rig wizards and ≈210 clicks vs 0
The concrete problems:
- Meshy's rigger refused the 903,526-triangle model Meshy had generated one message earlier, 3.0× over its own cap.
- The default wrist markers were wrong: one on the shoulder, one outside the body.
- Symmetry ON by default mirrored my left-wrist correction into a worse spot, so I switched it off and dragged three markers by hand.
For an animated brand mascot from one chat, I would use Snapshapy.
903,526 triangles meet a 300,000-face rigger
The model Meshy's rigger later refused came from Meshy's own approval card in message 2: Image to 3D, Meshy 7.1, Standard, texture 4096, Est. Cost 30, or $0.60 at the Pro rate of $0.02 a credit.
Meshy's image-to-3D approval card: Meshy 7.1, Standard, texture 4096, Est. Cost 30.
Meshy was quicker on this step, ≈92 s against Snapshapy's 291.7 s, and quicker end to end, ≈305 s against ≈475 s. The static GLB was 903,526 triangles and 44.5 MB with 3 JPEG textures.
My offline render of Meshy's static model, 903,526 triangles.
The Meshy model also kept fine branding details, a "Barista Bot" logo and a pressure gauge on the chest.
Meshy static model, front: the Barista Bot logo and the gauge.
Then I typed "Rig it and make it wave and dance." About 9 s later Meshy answered "Rigging failed": "The model currently has about 903,000 faces, which exceeds the 300,000 face limit for rigging."
- Meshy's model vs Meshy's rig limit — 903,526 vs 300,000 · 3.0× over
- Snapshapy's model vs its rig target — 1,500,000 vs 100,000 · 15× over · reduced without a card
The "Remesh to 10,000 faces" card, then "Rig Model"
The failure message arrived with a choice card whose first option was "Remesh to 10,000 faces". I selected it and clicked Submit: clicks 1 and 2 of message 3. The remesh took up to 112 s.
Meshy: rigging failed at about 903,000 faces vs a 300,000 limit, with the remesh choice card below.
A 10,000-face remesh is 30× below the limit Meshy had just quoted and roughly 90× below the model Meshy had built, and that size decision was mine. Then a "Rig Model" confirm card asked again what "rig it" had already said. Click 3. Meshy charged 0 credits for the remesh, the rig and both animations, so the balance stayed at 856.
A coffee robot with a 1.7 m default height
The rig wizard opened with a tip and a "Got it" button. Click 4. Step 1 had rotate, offset and height, with the height pre-filled at 1.7 m.
Meshy rig wizard step 1: rotate, offset and a default height of 1.7 m.
None of my messages gave the robot a size, so 1.7 m is a number the wizard brought and the user has to judge. I left all three controls alone and clicked Next. Click 5, and not one marker had appeared yet.
One wrist on the shoulder, one outside the body
Step 2 showed the robot with 13 markers already placed. The wizard guesses, the user corrects, and on this model the guesses failed exactly where the request needed them most.
Meshy's 13 default markers: one wrist marker on the shoulder, the other outside the body.
- Default markers — Meshy 13 placed, 2 wrong (15%) · Snapshapy 0 to place
- Markers I eventually moved — 3 of 13 (both wrists, one knee)
A wave is a hand motion, and the two hand markers were the two the wizard got wrong. Clicking Confirm without looking would have rigged the wave from a shoulder and a point in the air, with no warning.
Symmetry ON: my left-wrist correction pushed the other wrist further out
Symmetry was on by default. I dragged the left wrist marker into place, the wizard mirrored that move, and the right wrist marker landed in a worse spot than its default.
With Symmetry on, correcting one wrist mirrored the other further off the body.
The default meant to save a drag cost a click to switch it off, and the untouched wrist still needed its own drag:
- Symmetry off — 1 click
- Marker drags — 3 (left wrist, right wrist, one knee)
- Confirm — 1 click, then a rig job of ≈39 s
By Confirm, message 3 stood at 10 clicks and drags on Meshy.
Two pages of animations and "Approve 2 selected"
Next came a two-page choice card. Page 1, "Which waving animation": I picked Wave One Hand and clicked Next. Page 2, "Which dance animation" (Shake It Off Dance, Bubble Dance, Joyful Sway Dance, Custom): I picked Shake It Off Dance and clicked Submit.
Meshy's two-page animation picker, page 2: which dance.
Then "2 paid actions need your approval", and I clicked "Approve 2 selected". The two animation jobs took ≈27 s.
"2 paid actions need your approval" for the wave and the dance.
- Remesh card and "Rig Model" confirm — Meshy 3 · Snapshapy 0
- Wizard buttons, Symmetry off, marker drags — Meshy 7 · Snapshapy 0
- Animation picker and approval — Meshy 5 · Snapshapy 0
- Message 3 total — Meshy ≈15 · Snapshapy 0
With the cards and Approve clicks from the first two messages, the run reaches ≈21 clicks on Meshy against 0.
Snapshapy's message 3: "reduced to 100k for rigging · animated automatically"
Snapshapy answered the same sentence in one reply: "Rigging the robot now (3 credits, 1-3 minutes) … it will automatically get five clips: Greeting (a wave, 4.8 s) plus Dance Simple, Dance Body Roll, Dance Charleston and Dance Reach Hip. All the animation runs on its own — no further step from you."
Snapshapy's reply to "rig it and make it wave and dance": five clips will apply automatically, no further step.
About 11 s after the rig card, the animation card "5 clips · Greeting, Dance Simple…" appeared by itself, and the header read "14 spent", matching the balance drop from 1,021 to 1,007.
Snapshapy rig card ("reduced to 100k for rigging · animated automatically") and the 5-clip animation card; header "14 spent".
- Face-count problem — Meshy refused, user picks a remesh · Snapshapy reduced it inside the rig job
- Rig + animation jobs — Meshy ≈39 s + ≈27 s after the wizard · Snapshapy 118.8 s + 7.5 s, automatic
- Markers and wizard screens — Meshy 13 and 2 · Snapshapy 0 and 0
One GLB, a wave and four dances
Snapshapy's animated file held five clips: Greeting 4.83 s, Dance Simple 1.25 s, Dance Body Roll 6.5 s, Dance Charleston 2.33 s and Dance Reach Hip 2.53 s.
Frames I posed offline from Snapshapy's animated GLB, two per clip: Greeting and four dances.
- Files for wave + dance — Meshy 2 (27.1 MB + 27.4 MB = 54.5 MB) · Snapshapy 1 (47.4 MB)
- Dance clips — Meshy 1 · Snapshapy 4
Meshy's single dance is longer: Shake It Off Dance runs 16.3 s, longer than any Snapshapy dance clip (the longest is Dance Body Roll at 6.5 s, and all four add up to 12.61 s). Snapshapy's option 1 has no legs, just a pole on a round base, and in 3 of the 4 dances that pole and base stretch and twist in my offline pose frames. For store screens and social posts, four dances to rotate beat one.
The rigged meshes up close: 10,321 vs 99,993 triangles
Meshy rigged model wireframe, 10,321 triangles.
Meshy's rigged model is far lighter, 10,321 triangles against 99,993 in Snapshapy's 47.4 MB file with three 4096 px textures, and I never asked for a mobile version nor did Snapshapy offer one.
Wireframe of Snapshapy's animated GLB, 99,993 triangles.
- Reduction before rigging — Meshy 903,526 → 10,321, a 98.9% cut I chose on a card · Snapshapy 1,500,000 → 99,993, a 93.3% cut done in the job
- Hand work to get a rigged result — Meshy 10 clicks and drags · Snapshapy 0
Ten coffee robots: ten wizards, ≈210 clicks
Mascots rarely stay single. Message 1 already showed range on Snapshapy: a kettle-head barista on a pedestal, a mug-bodied bot, an espresso-machine bot and a ceramic pour-pot bot with a visor, in one cream, brown and copper palette.
Snapshapy's four concepts at full resolution: kettle-head barista on a pedestal, mug bot, espresso-machine bot, ceramic pour-pot bot.
Ten variants through the same three messages:
- Rig wizards — Meshy 10 · Snapshapy 0
- Clicks — Meshy ≈210 · Snapshapy 0
- Markers to inspect — Meshy 130 · Snapshapy 0
- Dance clips — Meshy 10 · Snapshapy 40
- Credits — Meshy 390 ($7.80) · Snapshapy 140 ($6.65)
My offline render of Snapshapy's static model, 1,500,000 triangles.
Final verdict: Meshy 3D Agent scores 53/100 for an animated brand mascot
That is lower than the 57 I opened with, because the rig step is the step a mascot exists for. Meshy's rigger refused the model Meshy itself had made, then asked me to pick a remesh, confirm the rig, accept a 1.7 m height, and repair 2 of 13 default markers at exactly the wrists a wave depends on. Symmetry made the first correction worse. Message 3 cost ≈15 clicks and drags, the run ≈21, and ten variants ≈210 with 10 wizards.
Which one I would use for this job
For an animated brand mascot from one chat, I would use Snapshapy. Its rig job did the face reduction itself, put Greeting and four dance clips in one 47.4 MB file, and asked for zero markers and zero clicks after "rig it". The plans are at snapshapy.com.