Facetra vs Kerf

Facetra is a browser-based AI jewellery design workflow for Indian ateliers and independent designers. It turns hand sketches, reference images, and CAD screenshots into photorealistic renders, then creates silhouette-locked views at 0°, 45°, 90°, and 180°. Supported Indian styles include Jadau, Polki, Kundan, Meenakari, Temple, and Rajwadi. Meshes can be exported as GLB, STL, OBJ, FBX, USDZ, or 3MF files. Additional tools include a searchable catalog, bulk PhotoRoom touch-up, lasso refinement, and a live 22kt gold-rate Bill of Materials estimator. Facetra can render screenshots from Rhino, MatrixGold, and JewelCAD. The maker positions generated meshes as references for CAD operators to retopologize before casting, not replacements for professional CAD software. New accounts receive 10 free credits, while purchased credits never expire. Plans include Starter at ₹1,500 one-time, Team at ₹4,800/month, Studio at ₹7,000 one-time, Atelier at ₹15,600 one-time, and Enterprise at ₹15,600/month.

Kerf is a free, open-source CAD workspace that runs on the user's own machine and supports work across 37 engineering domains. Users can describe design changes in plain language or use the Python kerf-sdk and JSON-RPC API. Mechanical design tools include constraint-based sketching, parametric feature history, validated B-rep geometry, drawings, and persistent face references. The workspace also includes FEM analysis, topology optimization, SPICE and RF analysis, CAM toolpaths, and G-code output. Listed outputs include STEP, STL, 3MF, IFC4, Gerber, PDF, DXF, BOM CSV, Touchstone, GDS-II, and firmware binaries. The site describes workflows for mechanical and electronics engineers, BIM architects, jewelry designers, and other engineering users. Kerf runs locally without accounts and keeps Git history for file revisions. Workshop content can be used offline once pinned. Installation options include source or Docker, with release archives listed for macOS arm64, macOS x64, and Linux x64; source installation requires Python 3.11 or later. The chat workflow can use a bring-your-own LLM; no bundled model or paid hosted plan is listed.