Blender Character — Near-Infrared (NIR) Skin, Eyes & Modular Accessories for Synthetic Face Dataset -- 2

via Freelancer ·

Budget / Salary₹12,500–37,500
TypeFreelance project
LocationRemote
Posted22 minutes ago
OVERVIEW
We are building a synthetic human face dataset in Blender using the open-source Google GNM parametric head extension:
https://www.reddit.com/r/blender/comments/1uzaxc3/google_gnm_open_source_parametric_head_released/

Please note: We will not be sending pre-made .blend files upfront. You can download and install the free Google GNM Blender extension directly and build all materials, eye shaders, hair, and accessories around the standard GNM head setup. Everything you deliver must plug cleanly and easily into our existing Google GNM Blender workflow.

Our goal is to use an automated Python rendering pipeline to randomize face shapes, skin profiles, hairstyles, facial hair, glasses, and headwear—generating diverse synthetic human portraits (similar in variety and framing to Microsoft’s FaceSynthetics dataset: https://github.com/microsoft/FaceSynthetics/blob/main/docs/img/dataset_samples.jpg), rendered to match real Near-Infrared (NIR) camera imagery.

Target IR Output & Eyeball References:
- Sample IR Face Output: https://link.springer.com/article/10.1007/s11263-021-01432-4
- High-Resolution NIR Eye & Iris Reference: https://www.researchgate.net/figure/High-resolution-NIR-face-images-and-corresponding-face-and-iris-images-a-High_fig1_221383633

(These links—along with additional sample IR images we will share with you—serve as our visual target. We will iterate together on your initial sample renders to fine-tune the look so it closely matches our actual camera output.)

We are looking for an experienced Blender look-dev / character artist to deliver:
1. An IR-accurate skin, eye, and mouth material library tailored to the Google GNM head.
2. A modular library of hair and accessories that fit cleanly across Google GNM head variations.
3. A calibrated lighting/render setup inside Blender that reproduces real NIR image characteristics.

------------------------------------------------------------

1. IR-ACCURATE SKIN, EYE & MOUTH MATERIAL LIBRARY
- IR Skin Tones & Ages: Distinct skin profiles covering very fair to very deep tones (referencing the Monk Skin Tone Scale) plus age variations (smoother skin for younger faces, fine lines/pores/sunspots for older faces), calibrated specifically for how melanin, subsurface scattering, and skin reflectance behave in monochrome Near-Infrared imagery.
- Close-Up Detail & Visible Skin Marks: High-resolution (2K minimum) PBR textures that hold up in tight portrait close-ups—ensuring pores, fine lines, and natural skin marks (moles, freckles, blemishes, scars, and light stubble shadow) remain clearly and accurately visible under IR illumination, with zero seams or stretching on the GNM head mesh.
- Script-Friendly Master Shader: A single master skin shader with clearly exposed, easy-to-script controls for skin tone/IR reflectance, roughness/shine, subsurface scattering, freckle/mole/blemish intensity, and stubble shadow, alongside pre-configured presets for all profiles.
- IR Eyeball & Iris Physics (Priority):
* Eyeballs must accurately replicate real Near-Infrared optical behavior (matching the NIR eye reference link above):
- Deep black pupil with a sharp white glint/dot (corneal specular reflection) when looking toward the camera.
- Clearly defined grey iris with realistic iris pattern/texture and natural IR reflectance across different iris shades.
- Bright white sclera with realistic wetness and a subtle tear line.
* Natural IR-tuned materials for lips, eyebrows, eyelashes, and interior mouth parts (teeth, gums, tongue) on the GNM mesh.

------------------------------------------------------------

2. MODULAR ACCESSORY LIBRARY (COMPATIBLE WITH GOOGLE GNM HEADS)
We need a comprehensive library of modular, mix-and-match assets built at real-world scale with clean topology and PBR materials, designed to maximize visual variety and generate a wide range of unique synthetic combinations:
- Head Hair & Eyebrows: A wide variety of hairstyles (short, medium, long, straight, wavy, curly, coily, tied back, shaved/buzz, etc.) plus diverse eyebrow shapes and densities in multiple natural shades (including grey and white).
- Facial Hair: A broad selection of styles (light and heavy stubble, short beards, full beards, moustaches, goatees, sideburns, etc.) across multiple shades.
- Eyewear: A diverse collection of prescription glasses (thin metal, thick acetate, rimless, half-rim) with clear refractive lenses, alongside sunglasses (aviator, square, round, wrap-around, oversized) with tinted, gradient, and mirrored lenses that behave realistically in IR.
- Headwear & Face Coverings: A wide assortment of headwear options (caps, beanies, wide-brim hats, headscarves, turbans, hoods, etc.) plus various surgical and cloth face masks.

GNM Compatibility & Automation Requirements:
- Parametric Fit: Because Google GNM generates varying face shapes, head sizes, and expressions, accessories must adapt/fit cleanly to generated GNM heads with zero clipping or floating gaps (e.g., bound via surface/mesh deformers or rig/shape-key compatible setups so they follow head shape changes automatically).
- Clean Layering: Accessories must combine together (for example, wearing glasses, a cap, and a beard simultaneously) without intersecting each other.
- Automation-Ready Structure: Because we will randomize combinations via a Python script, all objects, collections, materials, and shader parameters must follow a strict, consistent naming convention and clean hierarchy.

------------------------------------------------------------

TECHNICAL & LICENSING REQUIREMENTS
- Built and tested in Blender (Cycles) using native Blender tools and the free Google GNM extension—no paid add-ons required to open or render.
- Textures saved with relative paths or packed cleanly into the .blend files.
- Usage Rights: All work must be original or properly licensed so we have unrestricted rights to use the 3D assets internally to generate and use synthetic 2D image datasets commercially.

------------------------------------------------------------

DELIVERABLES
1. Master .blend file(s) built on the Google GNM head setup containing the IR-calibrated skin/eye/mouth shaders, lighting/render setup, and sample generated head variations.
2. Organized skin/eye material library and texture folders.
3. Modular accessory library .blend file(s) cleanly structured for Python script swapping and automatic fitting to GNM head shapes.
4. Brief documentation (README.txt) explaining how to apply everything into our Google GNM extension setup, naming conventions, and exposed shader parameters.

------------------------------------------------------------

HOW TO APPLY
Please include:
1. Portfolio samples showcasing your Blender look-dev, shading, or 3D asset work (2–3 close-up portrait renders showing realistic skin, eyes, hair/facial hair, or eyewear are given priority if available, though candidates with strong relevant 3D modeling/texturing skills can also apply).
2. A short note on how you will approach the IR skin marks, IR eyeball/iris reflection behavior, and fitting modular accessories to the parametric Google GNM head.
3. Your estimated timeline and cost broken down by milestones—specifically quoting an initial paid milestone batch consisting of IR skin profiles, the IR eyeball setup, 8–10 assets per accessory category, and initial sample renders for iterative alignment with our reference IR images. Once we review and approve this first batch (and confirm any adjustments needed), we will proceed with expanding the full asset library. Please include pricing both for this initial milestone batch and for adding more assets in subsequent milestones.
3d rendering blender 3d visualization blender 3d character design texturing
Apply on Freelancer →

Project sourced from Freelancer.com. Applications happen directly on the original platform — we never collect your data.