RFP for 3D, Static, Avatar Model

ObEN is seeking a professional service to create a 3D model of a celebrity’s face and full body (including head model, a hair model, and an outfit), based on 3D scans of the person’s face and body.  The service provider will also need to do the facial and body rigging according to ObEN’s specifications to be consistent with its rendering and animation tools.

  1. Clothing will be separate from the celebrity’s body scan.
  2. Hair: short, black, straight
  3. Input: high quality 3D scan of a person’s face, full body 3D scan, and one complete outfit captured in a professional studio:
    • 3D Scans – See reference section “ObEN Scan & Photo Requirements”
  4. Output: 3D static model
    • Blend shapes : 25 facial blendshapes are required and are listed in the image below. We are currently using 25 blendshapes for the avatar. 10 can be combined into 5 (eyeball movements): look left, look right, look up, look down, and blink.

    • Shaders: None
    • Model: All export models should follow default model topology. Eyelashes, hair should be hair cards.
    • Texture Maps:
      1. Renders will be baked into the diffuse channel before importing into the engine. Use as many or little textures maps required by the shader in order to achieve a high quality photorealistic skin affect. Maps expected include: subsurface scattering, normal, specular/metallic, gloss/roughness, diffuse, etc. Please provide the render scene with all materials set up for rendering.
      2. Hair should be made of hair cards using diffuse, normal, specular, and alpha. If the hair texture was made by Xgen, please provide FallOff, zDepth, and RBG render pass maps.
      3.  Clothing textures should be compatible with physically based rendering (PBR) and include albedo, metallic, and roughness.
      4. For file type, .PNG files are encouraged (.JPG’s will be sent back). Texture map resolution should be sized at 4096×4096. In case adjustments must be made, please also provide the original scene/source files.
    • Rig type: Blendshape-based for face rig and Joint-based for body rig
    • File Format options

      1. .ma file is preferred (Autodesk Maya Scene). If not,
      2. .fbx with the blendshapes connected. If none of those can be used,
      3. Obj’s of every mesh.
    • Rendering target environment: Unity
  1. Expected project start date 9/10/2019
  2. Turn-around time 7-10 days
  3. Art Style: Japanese with a mix of photorealistic and manga. Samples will be provided.

ObEN Scan & Photo Requirements

General Instructions

  • Please contact a professional 3D studio for data capture. High-end camera array (photogrammetry) setting or high resolution 3D scanner devices are recommended. Please share some sample data before the real capture.
    • Please wear tight clothes while scanning, we don’t need to include much clothes details in the full body scan, as we will produce the outfit separately with the body. Would be better to have quarts mesh, instead of triangles. But it’s fine if you cannot.
  • Make sure the face is always under a scattered and uniform lighting condition. Please avoid any colored lighting (white light is preferred). Wipe the face or put on some non-reflective powder if too oily to reduce highlight effect.

  • The entire face and forehead need to be always clearly visible. Pull up hair from forehead or to the back of ears, or use a headband (ref1, ref2) & remove framed glasses if necessary (contact lens are fine). Please also avoid shiny/reflective makeup (e.g., lip gloss, eye shadow), and take off accessories such as earrings.
  • If any specific wardrobe is requested for modeling, please wear the exact clothes for capturing. Otherwise, please try to wear tight clothes to show the body shape, with color as neutral as possible (e.g., grey, try to avoid too dark colors like black, or too bright colors like white). Please also try to avoid highly reflective material (e.g., shiny leather jacket).

3D Scans of Faces & Body

  • Sample 3D scan of the face looks like the follow (texture, geometry, wireframe)

  • Technical Specs
    1. Polygon Count: Ideally above 100,000 for the facial area. 
    2. Accuracy: Ideally the scanning error for each point is below 0.1 mm.
    3. Format: Ideally in the form of obj with mtl and texture file (jpg, png, etc.). Please provide the highest accuracy data possible, and DO NOT downsample or sub-divide the model when cleaning up.
  • Capture Reference
    • Please follow the reference to capture 1 body pose (A pose) and 4 facial expressions (neutral, natural closed mouth smile, mouth open, & “OOO”).
    1. For the neutral face, make sure to capture with mouth closed. For the smiley face, make sure it’s as natural as possible.
    2. If using Arctec (e.g, EVA, Spider) or other scanners that casts strong light, please capture with the light switched off to allow eyes open & various facial expressions.
    3. If using any handheld scanning device, please make sure the full face is captured at a optimized distance (close enough for higher resolution).

2D Photos of Neutral Face

  • Technical Specs
    1. Resolution: Ideally 4K, the face should take up as much area as possible.
    2. Lighting: uniformed, avoid too dark/bright photos & specular highlights.
    3. Perspective: as “flat” as possible (recommended to take photos at a distance and zoom in with the lens, should not sacrifice resolution).
    4. Format: Ideally raw image.
  • Capture Reference
    1. Please follow the reference to capture 8 + 3 photos. Remain a consistent neutral face during photo taking (this neutral face should also be consistent with the neutral face taken for 3D scan). Starting from front view, take a photo every 45 degrees (recommended that the participant to be seated in a rotating chair for easy execution), and take 8 pics in total. In addition, please take 3 photos of head up, head down and eyes closed.
    2. For the neutral face, make sure to capture with mouth closed.
    3. Please make sure the full face is captured at a optimized distance (close enough for higher resolution)

ObEN is an artificial intelligence company that creates complete virtual identities for consumers and celebrities in the emerging digital world. ObEN provides Personal AI that simulates a person’s voice, face and personality, enabling never before possible social and virtual interactions. Founded in 2014, ObEN is a Softbank Ventures Korea and HTC Vive X portfolio company and is located at Idealab in Pasadena, California.