Multi-view 3D generation works best when you shoot 8–12 photos in a 360° arc, keep the subject still, and light it evenly—then let the AI fuse those views into one clean mesh. That single rule prevents most bad outputs. Below is the practical playbook: exact angles, photo counts, lighting, and a step-by-step workflow from capture to STL.
Why multi-view beats single-photo generation
A single photo gives the AI one surface. It has to guess the back of the head, the far side of a jacket, and the depth of a nose. Multi-view images for 3D solve that by showing the same subject from several angles, so the model gets real geometry instead of guesses.
In practice, multi-view 3D generation reduces the two most common failures: warped silhouettes and melted details. When you feed a front, 3/4, side, and back view, the generator can cross-check proportions. That matters for figurines because a 1 mm error on a face becomes obvious at 15 cm tall.
If you only have one photo, start with the selfie to 3D figurine guide. If you can take more, this article is the better path.
The capture setup that actually works
You do not need a studio. You need consistency.
Camera and settings
- Use a phone or camera with a fixed focal length if possible. Avoid ultra-wide lenses; they bend edges and confuse the solver.
- Lock exposure and focus. On a phone, tap and hold the subject, then drag the exposure slider down slightly to protect highlights.
- Shoot at the highest resolution you have (12 MP or more). More pixels help fine details like glasses and hair strands.
- Keep the subject at the same distance for every shot. A 1.5–2 m distance works for a full-body figurine.
Lighting
- Use soft, even light: open shade, an overcast day, or two diffused lamps at 45° left and right.
- Avoid direct sun and harsh on-camera flash. Hard shadows read as geometry and create lumps.
- No backlighting. If the background is brighter than the subject, the silhouette gets clipped.
Background
- Plain, matte, and contrasting. A light gray wall or a bedsheet works.
- Avoid busy patterns, mirrors, and reflective floors. They add false features.
Best angles for 3D scanning (and how many photos)
For a figurine, the goal is coverage of the head, torso, arms, and legs—not a full object scan. Here is a reliable shot list.
The 8-shot core arc
Walk a full circle around the subject and stop at:
- Front (0°)
- Front-right 3/4 (45°)
- Right profile (90°)
- Back-right 3/4 (135°)
- Back (180°)
- Back-left 3/4 (225°)
- Left profile (270°)
- Front-left 3/4 (315°)
This is the minimum for multi-view 3D generation. It gives the solver a view of every major surface.
When to add more
- Add 4 more shots at 22.5° offsets if the subject has complex hair, a beard, or a detailed costume.
- Add a low angle (camera at knee height, tilted up) and a high angle (camera at eye height, tilted down) to capture the top of the head and the underside of the chin.
- For pets, shoot 12–16 photos because fur and ears create ambiguous silhouettes. See the pet 3D figurines guide for animal-specific tips.
How many photos for a 3D model?
- 8–12 photos: clean, simple subjects (single person, plain clothing).
- 12–20 photos: detailed subjects (layered clothing, props, textured hair).
- 20–40 photos: full photogrammetry-style capture with top and bottom coverage.
More is not always better. Duplicate angles and blurry frames add noise. Aim for sharp, evenly spaced views.
Photogrammetry angles vs. AI multi-view: what changes
Classic photogrammetry needs 40–100+ overlapping photos and a static scene. AI multi-view 3D generation is more forgiving: it can work from 8–12 images and tolerate small subject movement.
But the principles are the same. Photogrammetry angles—consistent distance, 60–80% overlap, no motion blur—still improve AI results. If you are coming from a photogrammetry background, keep your capture discipline and reduce the count.
A step-by-step multi-view 3D workflow
Step 1: Plan the pose
Choose a pose that reads well in 3D. Arms slightly away from the body, feet apart, head straight. Avoid crossed arms and hands in pockets; they merge into blobs.
Step 2: Capture in one pass
Move around the subject, not the subject around the camera. Ask them to hold still for 30–60 seconds. Use a tripod or a friend to keep the camera level.
Step 3: Cull and sort
Delete blurry, overexposed, or duplicate shots. Rename files in order (01_front, 02_front-right, etc.) so you can upload them in sequence.
Step 4: Generate the model
Upload the set to 3D Figurines. The tool, built on Tencent Hunyuan3D, fuses the views into a single mesh and exports GLB. If you need a printable file, convert to STL.
Step 5: Inspect and fix
Open the GLB in a viewer or Blender. Check for holes, floating fragments, and inverted normals. For common defects, follow the fix bad 3D model generations checklist.
Step 6: Prepare for printing
- Convert GLB to STL. For slicer compatibility, see best 3D export formats for slicers.
- Scale the figurine. A desk figurine is typically 10–15 cm tall; a keychain is 4–6 cm.
- Add supports for overhangs (chin, nose, arms).
- Slice with a 0.2 mm layer height for a balance of detail and speed. Use 0.12 mm for faces.
- Print in PLA or resin. Resin gives sharper facial detail; PLA is easier for beginners.
Step 7: Post-process
Remove supports, sand lightly with 400–800 grit, and prime before painting. For a full walkthrough, read the 3D printing figurines guide.
Multi-view 3D tips that save hours
- Shoot the face first. The head is the most scrutinized area. Take 3 close-up views (front, 3/4, profile) in addition to the full-body arc.
- Keep the background out of frame. Crop tight in-camera so the solver focuses on the subject.
- Avoid shiny fabrics. Satin, sequins, and metal reflect light unpredictably. If unavoidable, shoot in diffuse light and expect some cleanup.
- Use a color checker or a gray card. It helps if you plan to color-match later, though it is optional for geometry.
- Batch your captures. If you are making multiple figurines, keep the same lighting and distance for every subject. Consistency speeds up your editing.
- Test with a simple subject first. A plain t-shirt and jeans will teach you the workflow before you attempt a wedding dress or a sports uniform.
Common mistakes and how to avoid them
- Too few angles. Eight is the floor. If the back of the head is missing, the AI invents it.
- Moving the subject. Even 2 cm of drift between shots creates ghosting. Use a chair or mark foot positions with tape.
- Zooming between shots. Changing focal length changes perspective. Move your feet, not the zoom.
- Ignoring the top of the head. Add a high angle shot. Bald spots and hats need it.
- Uploading out of order. Some tools use sequence as a hint. Sort files before upload.
When to use multi-view vs. single photo
Use multi-view when:
- The figurine is a gift and needs a recognizable likeness.
- The subject has complex hair, a beard, or a costume.
- You are printing at 15 cm or larger.
Use a single photo when:
- You only have one good image, such as an old family photo. See old photos to 3D figurines.
- The subject is a pet or a person who cannot pose.
- You are making a small keychain where fine detail matters less.
Quick reference: capture checklist
- 8–12 photos in a 360° arc
- Fixed distance, locked exposure and focus
- Soft, even light; no harsh shadows
- Plain background
- Sharp, non-blurry frames
- Files sorted in order
- Extra close-ups of the face
From photos to a printable figurine
Multi-view 3D generation is a capture problem before it is a software problem. Get the angles and lighting right, and the AI has enough data to build a clean mesh. Then it is a standard 3D printing workflow: GLB out, STL for the slicer, supports, print, paint.
If you are new to the printing side, start with the how to create 3D models from photos primer. If you are editing the mesh, see Blender editing 3D figurines.
Ready to try it? Turn a photo into a printable figurine with 3D Figurines—upload your multi-view set, generate the model, and export STL for your printer.
