[0.4.0] - 2026-06-24

🎯 From a single prompt to a finished product — Creative Lab lands: hand it a photo or a line of text and get back a print-ready figure, lamp, keychain, or fridge-magnet, concept → build in a single call.

✨ Model toolkits is now complete — cheap format conversion (incl. 3MF), real-world resize, and clean UV unwrap for external texturing — plus two new image models, nano-banana-2 and gpt-image-2.

Added

• New API endpoints documented across meshy-3d-generation, meshy-3d-printing, and meshy-openclaw (SKILL.md + reference.md):
  • POST /openapi/v1/convert (1 credit) — convert a model to other formats (glb/fbx/obj/usdz/blend/stl/3mf) without remeshing. Cheapest way to get an STL/3MF.
  • POST /openapi/v1/resize (1 credit) — rescale to a real-world size: exactly one of resize_height / resize_longest_side / auto_size, optional origin_at.
  • POST /openapi/v1/uv-unwrap (5 credits, GA) — generate fresh UVs for a GLB (≤ 40,000 faces, else 400 → remesh first); outputs a GLB "UV white model" for external texturing.
  • Creative Lab (POST /openapi/creative-lab/{product}/v1/prototype 6 credits → .../build 30 credits) for figure / lamp / keychain / fridge-magnet. Documented primarily in meshy-3d-printing (physical products) and in meshy-openclaw. Build requires an API-created prototype of the same product (web-app prototypes return 404).
• New generation parameters in the reference tables: hd_texture (4K base color, meshy-6/latest only), decimation_mode (1–4 adaptive polycount), auto_size + origin_at, alpha_thumbnail (RGBA preview → alpha_thumbnail_url), multi_view_thumbnails (image/multi-image only), and input_task_id chaining for image-to-3d / multi-image-to-3d off a text-to-image / image-to-image result.
• New 2D image models: text-to-image and image-to-image now list nano-banana-2 and gpt-image-2 alongside nano-banana / nano-banana-pro.
• Every task GET response now documents consumed_credits so agents can report real spend.

Changed

• Refine fact-fix: refine works with meshy-5, meshy-6, or latest (= Meshy 6); pick the same family as your preview for consistency. Removed contradictory claims (the old "meshy-6 previews do NOT support refine" / mismatch caveats). Refine is 10 credits regardless of model.
• 3MF fact-fix: 3MF is supported but NOT included by default — request it via target_formats: ["3mf"], the Convert API, or the Multi-Color Print API (always 3MF).
• Pricing sync to the authoritative price list: Text-to-Image (nano-banana 3 / nano-banana-2 6 / nano-banana-pro 9 / gpt-image-2 9), Image-to-Image (3 / 6 / 9 / 12), Convert 1, Resize 1, UV Unwrap 5, Creative Lab prototype 6 / build 30.
• Deprecations noted where the params appear: symmetry_mode no longer affects output; art_style is ignored by Meshy-6; is_a_t_pose superseded by pose_mode; meshy-4 is retired (returns 400).