Image to STL: Meshy 3D Agent's 15 Extra Clicks to Rig a Mobile Knight

Image to STL: Meshy 3D Agent's 15 Extra Clicks to Rig a Mobile Knight

Image to STL, Meshy 3D Agent, Snapshapy

An image to stl workflow can produce a static model, but a mobile game character also needs rigging and animation. This test follows those later steps in Meshy 3D Agent and compares them with Snapshapy.

My third message to Meshy 3D Agent was "Rig it and give me idle, walk and attack animations." Before anything moved, Meshy's rigger refused the 2.3-million-face knight Meshy itself had built, I picked a 15,000-face remesh, confirmed the rig, clicked through a wizard, dragged a crooked chin marker and approved idle on its own card. Across the run: 15 clicks beyond my three messages. Snapshapy's 3D Agent asked for zero.

I tested both agents on my own paid plans, Meshy Pro ($20 a month, 1,000 credits) at meshy.ai and Snapshapy Pro at snapshapy.com, one conversation each. I ran the Snapshapy side the next day, after an update; the Meshy numbers are from the first day, 2026-11-03. The three messages were identical and typed verbatim: four concept options for a stylized knight for a mobile RPG, option 1 as a textured 3D model, then a rig with idle, walk and attack. Whenever a card offered choices, I took the first option.

This piece covers the last mile: from static model to something a phone game can load. Two asides: Meshy delivered 1 concept of the 4 I asked for, and its run cost 39 credits against Snapshapy's 14.

My verdict in 60 seconds — Meshy 3D Agent 60/100, Snapshapy 90/100

Meshy 3D Agent can end up with a playable knight, but it makes you assemble the path to it by hand, one card at a time.

  • Clicks beyond the 3 messages — Meshy 2/10 · Snapshapy 10/10 — 15 vs 0
  • Rigging its own model — Meshy 2/10 · Snapshapy 9/10 — refused at 2.3M faces vs a 300,000 limit; Snapshapy reduced to 20k by itself and rigged in 53.1 s
  • Hand work inside the rig — Meshy 3/10 · Snapshapy 10/10 — 3 wizard buttons, 13 markers, 1 drag and 2 animation Approves vs none
  • Clips matched to the request — Meshy 5/10 · Snapshapy 9/10 — 4 clips including an unrequested Running vs exactly Idle A, Walk, Sword Attack
  • File size for mobile — Meshy 3/10 · Snapshapy 9/10 — 35.7 MB vs 6.91 MB animated file, 5.2× smaller
  • Credits for the run — Meshy 4/10 · Snapshapy 9/10 — 39 vs 14, 2.8×
  • Speed — Meshy 8/10 · Snapshapy 5/10 — image-to-3D 132 s vs 389.8 s
  • Triangle count — Meshy 9/10 · Snapshapy 8/10 — 15,491 vs 20,000 in the animated mesh

The concrete problems:

  • Meshy's rigger rejected the 2,309,182-triangle model Meshy had just generated, 7.7× over its own limit.
  • The way out was mine to choose: a "Remesh to 15,000 Faces" card, select plus Submit.
  • The wizard asked me to check 13 markers by hand, and the chin marker was wrong.
  • Idle needed its own Approve card after the attack already had one.
  • The finished animated file weighed 35.7 MB, 23.8 s on a 12 Mbps connection.

For a mobile RPG character, I would use Snapshapy.

The tally: what I clicked on Meshy that wasn't a message

Meshy's image-to-3D approval card: Meshy 7.1, texture on, 4096, Est. Cost 30. One of four Approve clicks in the run.

I kept a count of every click Meshy asked for that was not typing a message. Itemized:

  • Choice cards (select + Submit) — Meshy 3 cards, 6 clicks · Snapshapy 0
  • Approve buttons — Meshy 4 · Snapshapy 0
  • Rig confirm card — Meshy 1 · Snapshapy 0
  • Wizard buttons (Got it, Next, Confirm) — Meshy 3 · Snapshapy 0
  • Marker drags — Meshy 1 · Snapshapy 0
  • Total — Meshy 15 · Snapshapy 0

Most of those clicks pile up after message 3. The remesh card, the animation choice card, the rig confirm, all three wizard buttons, the chin drag and two of the four Approves belong to the rig stage: 11 clicks for one sentence. The other 4 were the archetype card and the Approves on the 9-credit concept and the 30-credit 3D model.

Snapshapy's message 2 reply simply said "3 credits, about 5–8 minutes", with no card under it. Its composer has a default spend limit of "Up to 10" credits, above which it waits for approval; nothing in this run crossed it. At 15 clicks per knight, Meshy's tally is 180 clicks for a 12-character month.

2.3 million faces against a 300,000-face rigger

Meshy's reply to "rig it": rigging failed, about 2.3 million faces vs a 300,000 limit.

Message 3 failed after 9 s. Meshy's reply: "about 2.3 million faces, while the rigging tool requires fewer than 300,000 (10–20k ideal for mobile)".

The model was the one Meshy built on its own approval card, Meshy 7.1 at 4096. Meshy's own 3D output was too dense for Meshy's own rigger, and the agent tried the rig before checking.

  • Static textured GLB — Meshy 2,309,182 tris, 91.4 MB · Snapshapy 1,500,000 tris, 90.9 MB
  • Meshy's rig limit vs Meshy's model — 300,000 vs 2,309,182 · 7.7× over
  • Meshy's own mobile target — 10–20k faces · the model was 115× to 231× that

Snapshapy's static model was dense too, at 1,500,000 triangles, and it never turned into my problem. Its rig card reads "reduced to 20k for rigging · animated automatically": a 75× reduction done without a single click, landing inside the same 10–20k mobile window Meshy quoted.

The "Remesh to 15,000 Faces" card

Meshy's choice card offering "Remesh to 15,000 Faces".

To recover, the agent offered a choice card: "Remesh to 15,000 Faces". Select, Submit, 2 clicks. The remesh ran about 114 s and came back at ~15,500 faces.

Meshy animated model wireframe after the remesh: 15,491 triangles.

That is a 149× cut, from 2,309,182 triangles to 15,491 in the animated file, inside Meshy's stated mobile window. The mesh is fine. The problem is who makes the call: I had asked for a rig in plain words, and Meshy handed the decision back to me. Across 12 knights a month, that card means 24 clicks and 12 remesh jobs on a step I never asked for.

Wizard step 1: rotate, offset, height

After the remesh came a "Rig Model" confirm card, 1 click. Then the wizard opened with a tip, dismissed with "Got it". Step 1 shows rotate, offset and height controls and a Next button.

Meshy rig wizard step 1: rotate, offset and height before Next.

Confirm card, Got it, Next: 3 clicks before the wizard reaches the knight's body, for a step Snapshapy never showed me. At 12 characters that is 36 clicks a month.

Thirteen markers and a crooked chin

Meshy rig wizard step 2: 13 markers to check by hand before Confirm.

Step 2 places 13 markers on the knight and asks you to check them before Confirm. I dragged one: the chin marker sat off, so I pulled it into place. Then Confirm, and the rig took 84 s.

This is where the Meshy rig stops being an agent task. A misplaced chin marker on a helmeted knight is caught only by inspecting all 13 dots, so every character needs a careful look. At 12 knights a month that is 156 markers to check by eye.

Snapshapy showed me no markers and no wizard. Its rig job took 53.1 s, including its own reduce to 20k, and the 4.8 s animation job started the moment the rig finished, with nothing for me to touch in between.

Attack first, then a second Approve for idle

Meshy asks which animation to do first: Left Slash, Double Combo, Combat Stance or Other.

I had asked for idle, walk and attack in one sentence. Meshy answered with a choice card, "Which animation first", offering Left Slash, Double Combo, Combat Stance or Other, followed by an Approve. The attack took 24 s.

Meshy's separate approval card for the idle animation.

Idle did not follow on its own. It came as a second Approve card and took 28 s. Walk and run arrived bundled with the rig, so I got a running clip I never requested. Meshy's rig and both animations cost 0 credits, while Snapshapy charged 3 credits for rig plus animation.

The final Meshy reply also exposed an internal number: "Combat Stance (Idle) animation (action ID: 89)". Two animation approvals per knight is 24 extra cards a month.

Snapshapy's message 3: "reduced to 20k" and three clips I never approved

Snapshapy's reply to "rig it": 3 credits, Idle A, Walk and Sword Attack attached automatically, reduced to about 20,000 triangles with 1024 px textures "for a mobile RPG".

Snapshapy's answer to the same sentence: "Rigging now (3 credits, 1–3 minutes), and it'll pick up Idle A, Walk and Sword Attack automatically". Then the line that matters for this job: "Since it's for a mobile RPG I reduced it to about 20,000 triangles with 1024 px textures, so it comes out game-light rather than a 50 MB file".

Snapshapy rig card ("reduced to 20k for rigging · animated automatically"), the 3-clip animation card, and "14 spent" in the header.

The "mobile RPG" came from my first message, two messages earlier, and the agent carried it into the rig without being reminded. The decision Meshy put on a card, Snapshapy made without asking. The animation card says "Idle A, Walk, Sword Attack" and "3 clips": the three motions I asked for. The header read "14 spent", matching the balance drop from 1,035 to 1,021.

  • Clicks during the rig stage — Meshy 11 · Snapshapy 0
  • Clips asked vs delivered — Meshy 3 asked, 4 in the bundle · Snapshapy 3 asked, 3 delivered
  • Rig failures — Meshy 1 · Snapshapy 0

35.7 MB against 6.91 MB: the knight a phone has to download

Meshy animated model, front: 15,491 triangles.

The files the Meshy agent produced hold one action each: the attack file is 35.1 MB and the idle file is 35.05 MB. Pull both from the chat and you are carrying 70.15 MB for two clips of one knight.

Meshy download dialog: Animation "All Added", Single file on.

Meshy's download dialog can pack all 4 clips into one 35.7 MB GLB once you set "Animation: All Added" and "Single file". That bundle lives in a dialog with two settings, on top of the 15 clicks; the agent's own chat files stay single-clip.

Snapshapy animated model, front: 20,000 triangles, 1024 px textures.

I exported Snapshapy's animated card as GLB and opened it: 20,000 triangles, three 1024 × 1024 PNG textures, 6.91 MB, three clips in one file.

  • Animated GLB — Meshy 35.7 MB · Snapshapy 6.91 MB · 5.2× smaller
  • Download on 12 Mbps — Meshy 23.8 s · Snapshapy 4.6 s
  • Clips in the file — Meshy Running 0.63 s · Walking 1.03 s · Combat Stance 1.67 s · Left Slash 3.17 s · Snapshapy Idle A 3.12 s · Walk 1.67 s · Sword Attack 1.9 s
  • Dialog settings needed for one file — Meshy 2 · Snapshapy 0

Meshy's animated mesh is the leaner one by triangle count, 15,491 against Snapshapy's 20,000. On a phone, the 28.79 MB gap in file weight is the number a player feels.

Twelve knights a month: 180 clicks and 12 wizards

Snapshapy animated model wireframe: 20,000 triangles.

A mobile RPG rarely stops at one knight. Scale this run to 12 characters a month:

  • Extra clicks — Meshy 180 · Snapshapy 0
  • Rig wizards — Meshy 12 · Snapshapy 0
  • Markers to check by hand — Meshy 156 · Snapshapy 0
  • Remesh decisions — Meshy 12 · Snapshapy 0
  • Separate animation Approves — Meshy 24 · Snapshapy 0
  • Animated files — Meshy 428.4 MB · Snapshapy 82.9 MB · 5.2×
  • Credits — Meshy 468 ($9.36) · Snapshapy 168 ($7.98) · 2.8×
  • Extra clicks per year — Meshy 2,160 · Snapshapy 0

Meshy's image-to-3D was quicker, 132 s against Snapshapy's 389.8 s, and the whole Meshy run took ≈417 s against ≈519.2 s. But Snapshapy needed no input after its three messages, while Meshy's 15 clicks kept me at the tab through every stage.

Final verdict: Meshy 3D Agent scores 55/100 for a mobile RPG character

I score Meshy lower than in the opening because the rig stage is the job. Meshy refused its own 2,309,182-triangle model, then handed me a remesh, a rig confirm, a three-button wizard, 13 markers, a chin to drag and two animation approvals: 15 clicks per knight, 180 a month, 12 wizards, at 39 credits against 14. At the end of all that, Meshy's animated knight still weighed 35.7 MB. Snapshapy reached a rigged knight with Idle A, Walk and Sword Attack in one 6.91 MB file with zero extra clicks.

Which one I would use for this job

For a mobile RPG character, I would use Snapshapy. Three messages gave me a rigged knight, the three clips I asked for and a file sized for a phone, 6.91 MB with 1024 px textures, with no card, wizard or approval in the way, for 14 credits. The plans are at snapshapy.com.