meshy-ai – OpenClaw Skill

---

name: meshy-ai description: "Use the Meshy.ai REST API to generate assets: (1) text-to-2d (Meshy Text to Image) and (2) image-to-3d, then download outputs locally. Use when the user wants Meshy generations, needs polling async tasks, and especially when they want the resulting OBJ saved to disk. Requires MESHY_API_KEY in the environment."

Meshy.ai

Generate Meshy assets via API and save outputs locally.

Setup

• Add env var: MESHY_API_KEY=msy-...
• Optional: MESHY_BASE_URL (defaults to https://api.meshy.ai)

Text → 2D (Text to Image)

Use scripts/text_to_image.py.

python3 skills/public/meshy-ai/scripts/text_to_image.py \
  --prompt "a cute robot mascot, flat vector style" \
  --out-dir ./meshy-out

• Downloads one or more images (if multi-view) into ./meshy-out/text-to-image_<taskId>_<slug>/.

Image → 3D (always save OBJ)

Use scripts/image_to_3d_obj.py.

Local image

python3 skills/public/meshy-ai/scripts/image_to_3d_obj.py \
  --image ./input.png \
  --out-dir ./meshy-out

Public URL

python3 skills/public/meshy-ai/scripts/image_to_3d_obj.py \
  --image-url "https://.../input.png" \
  --out-dir ./meshy-out

• Always downloads model.obj (and model.mtl if provided by Meshy) into ./meshy-out/image-to-3d_<taskId>_<slug>/.

Notes

• Meshy tasks are async: create → poll until status=SUCCEEDED → download URLs.
• API reference for this skill: references/api-notes.md.

meshy-ai

Small Python helpers for the Meshy.ai REST API. The scripts create async tasks, poll until completion, and download outputs locally.

Features

• Text-to-Image (2D) generation with optional multi-view output.
• Image-to-3D generation with OBJ (and optional MTL) download.
• Simple polling + download helpers in a shared client.

Requirements

• Python 3.8+ (no third-party dependencies)
• Meshy API key

Setup

Set the API key in your environment:

export MESHY_API_KEY=msy-...

Optional base URL override:

export MESHY_BASE_URL=https://api.meshy.ai

Usage

Text → Image

python3 scripts/text_to_image.py \
  --prompt "a cute robot mascot, flat vector style" \
  --out-dir ./meshy-out

Optional flags:

• --ai-model (default: nano-banana)
• --aspect-ratio (only when not using multi-view)
• --generate-multi-view
• --timeout (seconds; default 900)

Outputs are saved to:

./meshy-out/text-to-image_<taskId>_<slug>/image_1.png

Image → 3D (OBJ)

Local file:

python3 scripts/image_to_3d_obj.py \
  --image ./input.png \
  --out-dir ./meshy-out

Public URL:

python3 scripts/image_to_3d_obj.py \
  --image-url "https://example.com/input.png" \
  --out-dir ./meshy-out

Optional flags:

• --ai-model (default: latest)
• --model-type (standard | lowpoly)
• --topology (triangle | quad)
• --target-polycount (int)
• --should-texture / --no-texture
• --timeout (seconds; default 1800)

Outputs are saved to:

./meshy-out/image-to-3d_<taskId>_<slug>/model.obj
./meshy-out/image-to-3d_<taskId>_<slug>/model.mtl   # if provided

Notes

• Tasks are asynchronous: create → poll → download when status=SUCCEEDED.
• API docs: https://docs.meshy.ai/en