Official Capture Guidelines
To maintain the commercial-grade standards required by top VFX studios and AI producers, all datasets must adhere to strict environmental and technical specifications. Clean, consistent data is the product — every shot below shows exactly what we expect.
1. Technical Requirements (Every Image)
Format & Resolution
- Accepted formats: PNG, JPEG (95%+ quality)
- Aspect ratio: 2:3 portrait — every shot, every tier, same shape
- Minimum resolution: 3000 × 4500 (≈13.5MP)
- Recommended: 4000 × 6000 (24MP)
- Max file size: 35MB per image
Camera & Setup
- Focal Length: 50mm+ equivalent. Wide-angle/smartphone-ultrawide lenses distort facial geometry.
- Stabilization: Tripod required, no handheld motion blur.
- Lighting: Flat, even 3-point rig. No moody shadows.
Clearances
- Wear plain clothing.
- No visible logos, branded clothing or third-party artwork.
- Cover tattoos unless you have the tattoo artist's permission.
- No other people, reflections or identifiable property in frame (Tier 1 natural backgrounds especially).
2. Dataset Tiers & Required Shots
Tier 1: 360° Head Turnaround
6 PhotosNeutral Background: A single, solid, non-patterned wall or backdrop (white, light gray, or beige) — no wallpaper, artwork, patterns, or visible furniture. Stand 3–4 ft in front of the backdrop to avoid casting a hard shadow onto it. Lighting should fall evenly on both you and the backdrop — no strong shadow on one side of your face.
Front
Neutral expression, eyes forward, mouth closed, camera at eye level.
45° Left
Head turned 45° from center, same neutral expression.
90° Left (Profile)
Full side profile, ear visible.
45° Right
Head turned 45° from center, same neutral expression.
90° Right (Profile)
Full side profile, ear visible.
Back of Head
Directly behind, hair/head shape fully visible.
Tier 2: Full Body Chroma Rig
16 PhotosChroma Screen: A solid blue or green backdrop, lit independently and evenly from the subject to avoid uneven exposure or color spill. Stand 6–8 ft from the screen to prevent green/blue reflection ("spill") on skin, hair, or clothing. Avoid wearing colors that match the screen.
A-Pose — Front
Arms relaxed at sides, slight gap from torso. Head to toe visible.
A-Pose — 45° Front-Right
Same A-pose, camera moved 45° to the right.
A-Pose — Right Profile
Full right side profile.
A-Pose — 45° Back-Right
Same A-pose, camera moved to back-right.
A-Pose — Back
Directly behind.
A-Pose — 45° Back-Left
Same A-pose, camera moved to back-left.
A-Pose — Left Profile
Full left side profile.
A-Pose — 45° Front-Left
Same A-pose, camera moved to front-left.
T-Pose — Front
Arms straight out to sides, parallel to ground.
T-Pose — Back
Same T-pose, directly behind.
T-Pose — Left Profile
Same T-pose, full left profile.
T-Pose — Right Profile
Same T-pose, full right profile.
Face — Neutral
Matches the Tier 1 front expression. Close crop, face and shoulders only.
Face — Smile/Joy
Natural smile, same framing as neutral shot.
Left Hand
Flat, fingers spread, palm facing camera.
Right Hand
Flat, fingers spread, palm facing camera.
Tier 3: Complete Digital Twin
16 Photos + 10 AudioChroma Screen: A solid blue or green backdrop, lit independently and evenly from the subject to avoid uneven exposure or color spill. Stand 6–8 ft from the screen to prevent green/blue reflection ("spill") on skin, hair, or clothing. Avoid wearing colors that match the screen.
A-Pose — Front
Arms relaxed at sides, slight gap from torso. Head to toe visible.
A-Pose — 45° Front-Right
Same A-pose, camera moved 45° to the right.
A-Pose — Right Profile
Full right side profile.
A-Pose — 45° Back-Right
Same A-pose, camera moved to back-right.
A-Pose — Back
Directly behind.
A-Pose — 45° Back-Left
Same A-pose, camera moved to back-left.
A-Pose — Left Profile
Full left side profile.
A-Pose — 45° Front-Left
Same A-pose, camera moved to front-left.
T-Pose — Front
Arms straight out to sides, parallel to ground.
T-Pose — Back
Same T-pose, directly behind.
T-Pose — Left Profile
Same T-pose, full left profile.
T-Pose — Right Profile
Same T-pose, full right profile.
Face — Neutral
Matches the Tier 1 front expression. Close crop, face and shoulders only.
Face — Smile/Joy
Natural smile, same framing as neutral shot.
Left Hand
Flat, fingers spread, palm facing camera.
Right Hand
Flat, fingers spread, palm facing camera.
Plus: 10 Audio Recordings
8 phoneme-coverage takes reading Amsaya's provided script, plus 2 performance recordings (a natural 30-second monologue and one expressive read). Full script provided during upload.
Read Amsaya's script exactly as written. The script belongs to Amsaya.
- Formats: WAV (preferred) or MP3 (256kbps+)
- Max 100MB per file
- Quiet, sound-treated environment — no background noise or echo
- Language: English (initial launch)
Voice Only
No photos · 10 AudioA voice-only dataset is just your voice. You save it, then do a short live voice check and record ten scripts in your browser, in one sitting. A moderator listens to every recording before it is approved.
Read Amsaya's script exactly as written. The script belongs to Amsaya.
- Quiet, sound-treated environment: no background noise or echo
- Language: English (initial launch)
- You tell us whether it is your natural voice or a character or performed voice you created and perform yourself. Studios see that label.
- It is only your own voice, never another person's, and never one made or cloned by software.
- There is no Background Crowd Bundle for a voice-only dataset, because no image of you is included.
Ready to Capture?
Review the guidelines, set up your studio space, and sign up to start monetizing your first dataset.
Sign Up to Get Started