3D Depth Viewer(Supports 2D-to-3D conversion)

Ver 1.3 — Help & Reference (English)
🇯🇵 日本語

🎯 Overview

3D Depth Viewer is a browser-based RGBD(2D+Depth) to 3D image and video viewer. It supports Side-by-Side (SBS), anaglyphs, parallax barrier, interlaced formats, and more. It also features on-device AI depth estimation to convert any ordinary 2D photo into a 3D depth-based view. It seems that AI depth estimation on mobile devices is difficult.

No installation is needed — everything runs directly in your browser using WebGL and WebGPU.

💡 The viewer automatically hides the toolbar after a few seconds of inactivity. Move your mouse or tap the screen to bring it back.

🔧 Toolbar

The toolbar appears at the top of the screen. Hover (or tap) any menu to open its sub-menu.

Main Buttons

ButtonFunction
✕ (Close)Go back to the previous page
⏺ RecRecord the current 3D view to a video file
⛶ FullscreenToggle fullscreen mode
⊡ FitFit the image to the screen
⇄ L/RSwap left and right eye channels
↔ FRReverse depth direction (foreground ↔ background)
◀ / ▶Previous / Next image or video

Drop-down Menus

MenuFunction
Stereo ▾Select the 3D display format (see below)
Slide ▾Slideshow interval: Off / 3 / 5 / 7 / 10 sec
File ▾Load image or video files
Help ▾Open this help page, version info, reset settings

👓 Stereo Modes

Choose a mode that matches your display or glasses.

2D+Depth
Depth preview — original image with depth applied
SBS
Full-width Side-by-Side (for SBS headsets / TVs)
HSBS
Half-width SBS (for half-SBS displays)
LR/RL
Left-Right / Right-Left (cross-eyed or parallel free-view)
Dubois
Low ghost anaglyph (Red-Cyan glasses recommended)
Color
Color anaglyph (Red-Cyan glasses)
Gray
Grayscale anaglyph (Red-Cyan glasses)
H_Int
Horizontal interlaced (row-interleaved 3D displays)
3DLCD
Column-interleaved for autostereoscopic LCD panels
🔄 Cycle through all stereo modes by pressing S on the keyboard.

📂 File Loading

You can load files using the following methods. You can also load multiple files. When you load multiple files, you can navigate forward and backward between them and view them in a slideshow.

Via the File Menu

OptionDescription
Full SBSFull-width Side-by-Side RGBD image / video
Half SBSHalf-width Side-by-Side RGBD image / video
Full T/BFull-height Top-Bottom RGBD image / video
Half T/BHalf-height Top-Bottom RGBD image / video
AI 2D→3DConvert a standard 2D image to 3D using AI depth estimation

Drag & Drop

You can drag and drop image or video files directly onto the viewer window. If you drop multiple files at once, you can use the forward/backward buttons or view them in a slideshow. It will convert and display the file in the format last selected from the File menu.

🎛️ Depth & Focus Sliders

The slider panel appears at the bottom of the screen. It stays visible while you are interacting with it.

Depth (0–100)

Controls the stereoscopic depth (parallax amount). Higher values push the 3D effect further apart; lower values flatten it. The internal value is depth / 2. Use the keyboard arrow keys for fine adjustment. Press C or Home to reset.

Focus (0–255)

Adjusts the zero-parallax plane (convergence). Lower values move focus closer; higher values move it further away.

Each slider has a Reset button to return it to its default value.

▶️ Video Controls

Video controls are displayed in the toolbar only when a video is loaded. They are automatically hidden when viewing still images.

ButtonFunction
▶ Play / ⏸ PauseToggle video playback
⏹ StopStop and reset playback
🔊 Vol+Increase volume by 10%
🔉 Vol−Decrease volume by 10%
🔇 MuteToggle mute

The progress bar in the slider panel shows playback position. Click anywhere on the bar to seek. It is hidden when a still image is displayed.

⌨️ Keyboard Shortcuts

KeyAction
Space / ¥Next image / video
Backspace / ;Previous image / video
SCycle through stereo modes
P / Play / Pause video
EnterToggle fullscreen
FFit image to screen
HCycle image aspect ratio (SBS / T/B / normal)
WToggle depth reverse (foreground ↔ background)
XSwap left / right eye
/ Increase depth (parallax)
/ Decrease depth (parallax)
KRotate image clockwise
LRotate image counter-clockwise
C / HomeReset depth and rotation to defaults
MShow / hide toolbar & sliders
QQuit (go back in history)
Numpad 0–9Set slideshow interval (0 = off, 1–9 = seconds)

🖱️ Touch & Mouse Operations

Mouse

ActionEffect
Drag (left button)Pan the image
Scroll wheelZoom in / out
Mouse moveShow the toolbar (resets auto-hide timer)

Touch (Mobile / Tablet)

GestureEffect
Single finger dragPan the image
Pinch in / outZoom in / out
TapShow the toolbar

🤖 AI 2D→3D Conversion

The AI 2D→3D feature uses the Depth Anything V2 Small model (via Transformers.js) to estimate a depth map from any ordinary photograph and render it as a stereoscopic 3D image directly in your browser.

How to use

  1. Open File ▾ → AI 2D→3D and select a photo.
  2. First-time only: a confirmation dialog will appear warning that the model data (~100 GB in total ONNX + runtime cache) will be downloaded. Confirm to proceed.
  3. The model is downloaded and cached in your browser. Subsequent uses skip the download.
  4. The depth map is estimated and the image is displayed in the selected stereo mode.
⚠️
First-time download: approx. 100 GB
The model files are large. Ensure you have sufficient disk space and a fast internet connection before proceeding. The confirmation dialog appears only if the files have not been cached yet.

Performance

The viewer automatically uses WebGPU (GPU-accelerated) if available, and falls back to WASM / CPU mode on unsupported browsers. WebGPU is significantly faster and recommended for a smooth experience.

📦 Open Source Acknowledgements

3D Depth Viewer is built with the following outstanding open-source projects. We are deeply grateful to all contributors.

🔺
Three.js
MIT License
The core 3D rendering engine used for WebGL-based stereoscopic display, shader materials, and camera management.
https://threejs.org/
🤗
Transformers.js
Apache 2.0 License
Hugging Face's JavaScript library for running machine learning models directly in the browser via ONNX Runtime.
https://huggingface.co/docs/transformers.js
🔭
Depth Anything V2 Small
Apache 2.0 License
State-of-the-art monocular depth estimation model by the Depth Anything Team. Used to generate depth maps from ordinary 2D images for AI 2D→3D conversion.
https://depth-anything-v2.github.io/
Model: onnx-community/depth-anything-v2-small
📋
Each library is used in accordance with its respective license. Please refer to the individual project pages for full license texts. The Depth Anything V2 model weights and Transformers.js runtime are downloaded from the Hugging Face Hub and CDN respectively.