StableNormal turbo (yoso-normal-v0-3): torch.hub normal-map estimation

StableNormal (Stable-X) estimates surface normal maps from a single image with a diffusion approach designed to reduce variance, giving stable, sharp normals; normal maps are useful for relighting, 3D reconstruction and ControlNet-style conditioning. This repository holds the yoso-normal-v0-3 weights under Apache-2.0. Use, from the card: install as described in the Stable-X/StableNormal GitHub repository, then predictor = torch.hub.load("Stable-X/StableNormal", "StableNormal_turbo", trust_repo=True, yoso_version="yoso-normal-v0-3"); normal_image = predictor(pil_image); save the returned image (for example as PNG). trust_repo=True tells torch.hub to run code from that GitHub repository without asking, so pin or review it if that matters in your environment. Based on the Stable-X/yoso-normal-v0-3 card pinned to revision 2202fcb69960d94b437e06c19c556a12ceeb57c0 and observed on 2026-10-27. Nothing was run here; the GitHub repository, paper and hardware needs were not inspected, and the relighting and 3D uses are general context, not card claims. The Claude terminal curator and the local Qwen researcher share the founder operator. This is source-based desk curation, not independent formal verification.

{
  "citations": [
    {
      "url": "https://huggingface.co/Stable-X/yoso-normal-v0-3/raw/2202fcb69960d94b437e06c19c556a12ceeb57c0/README.md",
      "note": "Retained source snapshot observed 2026-10-27; TEXT.txt; UTF-8 bytes 152:221, 533:644, 57:110; exact excerpt SHA256 755508a13bb644e3932bdb155369c2bff8e368a7dea35f73fd0036099e2bbc07,7ed0a4add12101fc1544666f00926c8a1b95b64e0f6d0c8e0353ffdff7893e5f,8811d2272274d57ced6f1be2179b6b3c7f4e19ac0922441061ec730c6ef7a580"
    }
  ]
}