The All-in-One 3D Suite for Creators

This is Hybrid 3D.
Built from AI depth + custom stereo logic —
Designed for cinema in VR.

Click to download or support the project 💙

Official website out now →

---

Notice

VisionDepth3D is licensed under a proprietary, no-derivatives license.
Forking, redistributing, modifying, or creating derivative works is strictly prohibited.

---

Table of Contents

• Key Features
• Guide Sheet: Install
• Documentation
• Dev Notes
• Acknowledgments & Credits

All-in-One 3D Suite

3D Generator (Stereo Composer)

(3D Generator Tab)

• GPU-accelerated stereo warping: per-pixel, depth-aware parallax shifting (CUDA + PyTorch)
• Built on the VisionDepth3D Method, including:
  • Depth shaping (Pop Controls): percentile stretch + subject recenter + curve shaping for natural separation
  • Subject-anchored convergence: EMA-stabilized zero-parallax tracking for comfort and consistency
  • Dynamic stereo scaling (IPD): scene-aware intensity that adapts to depth variance
  • Edge-aware masking + feathering: suppress halos and clean up subject boundaries
  • Floating window (DFW): cinematic edge protection to prevent window violations
  • Occlusion healing: fills stereo gaps and reduces edge artifacts
• Live preview + diagnostics: anaglyph, SBS, heatmaps, edge/mask inspection, stereo difference views
• Clip-range rendering for fast testing on difficult scenes before full renders
• Export formats: Half-SBS, Full-SBS, VR (SBS 1440×1600 per eye), Anaglyph, Passive Interlaced
• Encoding pipeline: FFmpeg with CPU and hardware encoders (NVENC/AMF/QSV) plus quality controls (CRF/CQ)

Result: A production-ready 2D-to-3D engine with real-time tuning tools, stability features, and flexible export formats for VR and cinema workflows.

---

Depth Estimation (AI Depth Engine)

(Depth Estimation Tab)

• 25+ supported depth models (ZoeDepth, MiDaS, DPT/BEiT, DINOv2, DepthPro, Depth Anything V1/V2, Distill-Any-Depth, Marigold, and more)
• One-click model switching with auto-download + local caching
• Multiple inference backends:
  • PyTorch (Transformers / TorchHub)
  • ONNXRuntime (CUDA / TensorRT)
  • Diffusers FP16 (for diffusion-based depth)
• Image + video + batch workflows:
  • Single image
  • Image folder batch
  • Full video depth rendering
  • Video folder batch
• Optional high precision output (when supported) for cleaner disparity and stronger stability in post
• Built-in preview modes + colormaps for fast inspection
• Stability + safety features: resolution/shape handling, codec probing, and fallback behavior to avoid common crashes

Result: Fast, flexible depth generation for everything from quick tests to full-length depth videos ready for stereo conversion.

---

Depth Blender (Multi-Source Depth Fusion)

(Depth Blender Tab)

• Blend two depth sources into one cleaner, more stable depth map/video
• Frames or video mode:
  • Pair two PNG frame folders
  • Or pair two depth videos
• Live preview + scrubber: side-by-side (Base vs Blended) with fast frame navigation
• Edge-focused blend controls:
  • White strength injection
  • Feather blur smoothing
  • CLAHE contrast shaping
  • Bilateral edge-preserving denoise
• Normalization back to base for consistent depth scale
• Batch output options: overwrite base, output to new folder, or export a blended video

Result: Cleaner edges, stronger subject separation, and more consistent parallax behavior across full sequences.

---

FPS / Upscale Enhancer (RIFE + Real-ESRGAN)

(FPS / Upscale Enhancer Tab)

• RIFE interpolation (ONNX): 2× / 4× / 8× FPS generation with GPU acceleration
• Real-ESRGAN upscaling (ONNX): high-quality super-resolution with optional FP16
• Two processing pipelines:
  • Merged (stable, low memory)
  • Threaded (higher throughput, better utilization)
• Full video workflow support:
  • Optional scene splitting for long videos
  • Rebuild output with correct resolution, FPS, and encoding settings
• Render feedback: progress, FPS, ETA, logs, and safe cancel handling

Result: Turn low-res or low-FPS sources into clean, smooth outputs built for VR playback and high refresh displays.

---

Live 3D / Realtime Stereo

(Live 3D Tab)

The Live 3D tab brings realtime stereo conversion into VisionDepth3D. It allows users to capture a camera, capture card, or screen source, estimate depth live, and preview a stereoscopic 3D output without waiting for a full render.

• Realtime capture sources:

  • Camera / capture card input
  • Screen 1 / Screen 2 desktop capture
  • Configurable capture resolution and FPS

• Depth model selection:

  • Uses the same supported model list as the Depth Engine
  • Lightweight model defaults for live performance
  • Optional FP16 acceleration where supported

• Live VisionDepth3D stereo controls:

  • Foreground / midground / background shift
  • Max pixel shift
  • Parallax balance
  • Depth pop gamma
  • Subject tracking
  • Dynamic convergence
  • Edge masking
  • Feathering
  • Floating window support

• Preview and output controls:

  • Start directly in SBS mode
  • Adjustable preview resolution
  • Optional preview window disabling
  • Optional HTTP stream field for future streaming workflows

• Designed for fast tuning:

  • Test stereo settings before rendering
  • Check depth direction and pop-out behavior
  • Compare depth models quickly
  • Tune comfort settings before full video export

Result: A realtime VisionDepth3D sandbox for testing depth models, stereo settings, screen capture, and live 2D-to-3D conversion before committing to final renders.

---

Smart GUI + Workflow

• Multi-tab interface with persistent settings
• Help menu
• Pause, resume, and cancel for long GPU jobs
• Multi-language UI support (EN, FR, ES, DE, JA)
• Hardware encoding options integrated into export workflow

---

Output Formats & Aspect Ratios

• Stereo formats: Half-SBS, Full-SBS, VR180, Anaglyph, Passive Interlaced
• Aspect ratios: 16:9, 2.39:1, 2.76:1, 4:3, 21:9, 1:1, 2.35:1
• Containers: MP4, MKV, AVI
• Encoders: CPU + FFmpeg hardware options (NVENC/AMF/QSV) when available

Guide Sheet: Install

📌 System Requirements

• ✔️ Python 3.13
• ✔️ Git, if cloning the repository
• ✔️ Conda, optional but recommended
• ✔️ NVIDIA GPU recommended for best performance
• ✔️ CUDA 12.8 tested
• ✔️ AMD / Intel GPU support on Windows through DirectML
• ✔️ CPU fallback available, but much slower

---

📌 Step 1: Download or Clone VisionDepth3D

You can install VisionDepth3D in one of two ways:

Option A: Download ZIP

1. Go to the official VisionDepth3D GitHub page.
2. Click the green Code button.
3. Click Download ZIP.
4. Extract the ZIP to a folder, for example:

C:\VisionDepth3D-main

Option B: Clone with Git

Open Command Prompt or Anaconda Prompt and run:

git clone https://github.com/VisionDepth/VisionDepth3D.git
cd VisionDepth3D

If you downloaded the ZIP instead, skip git clone and use cd to enter the extracted folder:

cd C:\VisionDepth3D-main

---

📌 Step 2: Create Environment and Install Dependencies

🟢 Option 1: Standard pip Install

Open Command Prompt:

cd C:\VisionDepth3D-main
pip install -r requirements.txt

Then continue to Step 3: Install PyTorch for Your GPU.

---

🔵 Option 2: Conda Install, Recommended

Conda is recommended because it keeps VisionDepth3D dependencies isolated from the rest of your system.

Open Anaconda Prompt and run:

git clone https://github.com/VisionDepth/VisionDepth3D.git
cd VisionDepth3D
conda create -n VD3D python=3.13 -y
conda activate VD3D
pip install -r requirements.txt

If you downloaded the ZIP instead of cloning:

cd C:\VisionDepth3D-main
conda create -n VD3D python=3.13 -y
conda activate VD3D
pip install -r requirements.txt

---

📌 Step 3: Install PyTorch for Your GPU

VisionDepth3D uses PyTorch for AI depth models and GPU processing.

Different GPU types need different PyTorch installs.

---

🟩 NVIDIA GPU Users, CUDA Recommended

If you have an NVIDIA GPU, install the CUDA build of PyTorch.

First, check your NVIDIA driver/CUDA support:

nvidia-smi

You can also check CUDA Toolkit if installed:

nvcc --version

Then install PyTorch using the official PyTorch selector:

🔗 https://pytorch.org/get-started/locally/

Recommended selector options:

OS: Windows or Linux
Package: Pip
Language: Python
Compute Platform: CUDA

Example for CUDA 12.8:

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128

If your system uses a different CUDA version, use the command from the official PyTorch website instead. PyTorch’s install selector is the safest source for the correct command.

---

🟥 AMD / Intel GPU Users on Windows, DirectML

If you have an AMD GPU or Intel GPU on Windows, install PyTorch DirectML.

DirectML allows PyTorch acceleration on supported non-NVIDIA GPUs through Windows DirectX 12.

Run this inside your VisionDepth3D environment:

pip install torch-directml

Use this option for:

• AMD Radeon GPUs on Windows
• Intel Arc / Intel integrated GPUs on Windows
• Systems without NVIDIA CUDA support

Important:

• DirectML is usually slower than NVIDIA CUDA.
• Some models or operations may fall back to CPU.
• If DirectML gives issues, use CPU mode as a fallback.
• Do not install CUDA PyTorch for AMD GPUs on Windows.

---

⬜ CPU-Only Users

If you do not have a supported GPU, install the CPU version of PyTorch.

Use the official PyTorch selector:

🔗 https://pytorch.org/get-started/locally/

Recommended selector options:

OS: Windows / Linux / Mac
Package: Pip
Language: Python
Compute Platform: CPU

CPU mode works, but depth generation, upscaling, interpolation, and 3D processing will be much slower.

---

📌 Step 4: Verify PyTorch Install

After installing PyTorch, test it.

NVIDIA CUDA Test

python -c "import torch; print('CUDA:', torch.cuda.is_available()); print(torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'No CUDA GPU found')"

AMD / Intel DirectML Test

python -c "import torch_directml; d=torch_directml.device(); print('DirectML device:', d)"

CPU Test

python -c "import torch; print('PyTorch installed:', torch.__version__)"

---

📌 Step 5: Launch VisionDepth3D

After all dependencies are installed, launch VisionDepth3D with the correct script for your setup.

Windows Conda:

Start_VD3D_Conda.bat

Windows standard install:

Start_VD3D_Windows.bat

Linux:

Start_VD3D_Linux.bat

Or run directly:

python app.py

---

📌 Notes for Batch Scripts

If you are using Conda, make sure your batch script activates the correct environment:

conda activate VD3D
python app.py

If you are using standard pip without Conda, make sure Python is available in PATH:

python app.py

---

✅ Install Complete

Congrats, you have successfully installed VisionDepth3D.

Recommended setup:

• NVIDIA users: CUDA PyTorch
• AMD / Intel Windows users: torch-directml
• No GPU users: CPU PyTorch

For the best performance, an NVIDIA CUDA GPU is recommended.

---

🛠️ Guide Sheet: Updating VisionDepth3D

When a new version of VisionDepth3D is released, follow these steps to ensure a smooth transition:

Update Instructions

1. Backup Your Weights
   Move your weights folder out of the old VisionDepth3D-main directory.

2. Download the Latest Version
   Delete the old folder and extract or clone the updated version of VisionDepth3D-main.

3. Restore Weights Folder
   Place your weights folder back inside the newly downloaded main directory:
   VisionDepth3D-main/weights

4. Update the Path in Startup Scripts
   Open the startup script matching your platform:

   • Start_VD3D_Windows.bat
   • Start_VD3D_Conda.bat
   • Start_VD3D_Linux.sh

   Edit the script and replace any old folder path with the new path to your updated VisionDepth3D-main.

5. Activate Conda Environment (if needed)
   If you are using the Conda starter script:

   • Open a terminal or Anaconda Prompt.
   • Run:

     cd path/to/updated/VisionDepth3D-main
     Start_VD3D_Conda.bat

6. Launch the App
   Once everything is in place, run the appropriate script or shortcut to launch VisionDepth3D with your latest settings.

---

Note: If you customized any configuration, backup those files before replacing folders. and if you run into import errors

pip install -r requirements.txt

inside opened terminal and that will fix any dependancie errors

---

📘 Documentation

VisionDepth3D includes a full professional user manual with workflows, tuning guides, and advanced features.

👉 Start here: UserGuide.md

If you're new, begin with: • Depth Estimation →
• Depth Blender →
• 3D Generator →
• Preview & Clip Testing →
• Final Render

---

Dev Notes

This tool is being developed by a solo dev with nightly grind energy (🕐 ~4 hours a night). If you find it helpful, let me know — feedback, bug reports, and feature ideas are always welcome!

Acknowledgments & Credits

Thank You!

A heartfelt thank you to all the researchers, developers, and contributors behind the incredible depth estimation models and open-source tools used in this project. Your dedication, innovation, and generosity have made it possible to explore the frontiers of 3D rendering and video processing. Your work continues to inspire and empower developers like me to build transformative, creative applications.

Supported Depth Models

Model Name (UI)	Creator / Organization	Model ID / Repository
Marigold Depth v1.1 (Diffusers)	PRS ETH	prs-eth/marigold-depth-v1-1
Marigold Depth v1.0	PRS ETH	prs-eth/marigold-depth-v1-0
Distill-Any-Depth Large (xingyang1)	xingyang1	xingyang1/Distill-Any-Depth-Large-hf
Distill-Any-Depth Small (xingyang1)	xingyang1	xingyang1/Distill-Any-Depth-Small-hf
Video Depth Anything Large	Depth Anything Team	depth-anything/Video-Depth-Anything-Large
Video Depth Anything Small	Depth Anything Team	depth-anything/Video-Depth-Anything-Small
Video Depth Anything (ONNX)	Depth Anything Team	Bundled ONNX backend (onnx:VideoDepthAnything)
Distill-Any-Depth Large (ONNX)	xingyang1	Bundled ONNX backend (onnx:DistillAnyDepthLarge)
Distill-Any-Depth Base (ONNX)	xingyang1	Bundled ONNX backend (onnx:DistillAnyDepthBase)
Distill-Any-Depth Small (ONNX)	xingyang1	Bundled ONNX backend (onnx:DistillAnyDepthSmall)
DA3METRIC-LARGE	Depth Anything Team	depth-anything/DA3METRIC-LARGE
DA3MONO-LARGE	Depth Anything Team	depth-anything/DA3MONO-LARGE
DA3-LARGE	Depth Anything Team	depth-anything/DA3-LARGE
DA3-LARGE-1.1	Depth Anything Team	depth-anything/DA3-LARGE-1.1
DA3-BASE	Depth Anything Team	depth-anything/DA3-BASE
DA3-SMALL	Depth Anything Team	depth-anything/DA3-SMALL
DA3-GIANT	Depth Anything Team	depth-anything/DA3-GIANT
DA3-GIANT-1.1	Depth Anything Team	depth-anything/DA3-GIANT-1.1
DA3NESTED-GIANT-LARGE	Depth Anything Team	depth-anything/DA3NESTED-GIANT-LARGE
DA3NESTED-GIANT-LARGE-1.1	Depth Anything Team	depth-anything/DA3NESTED-GIANT-LARGE-1.1
Depth Anything v2 Large	Depth Anything Team	depth-anything/Depth-Anything-V2-Large-hf
Depth Anything v2 Base	Depth Anything Team	depth-anything/Depth-Anything-V2-Base-hf
Depth Anything v2 Small	Depth Anything Team	depth-anything/Depth-Anything-V2-Small-hf
Depth Anything v2 Metric Indoor (Large)	Depth Anything Team	depth-anything/Depth-Anything-V2-Metric-Indoor-Large-hf
Depth Anything v2 Metric Outdoor (Large)	Depth Anything Team	depth-anything/Depth-Anything-V2-Metric-Outdoor-Large-hf
Depth Anything v2 Giant (safetensors)	Depth Anything Team	Local weights (dav2:vitg_fp32)
Depth Anything v1 Large	LiheYoung	LiheYoung/depth-anything-large-hf
Depth Anything v1 Base	LiheYoung	LiheYoung/depth-anything-base-hf
Depth Anything v1 Small	LiheYoung	LiheYoung/depth-anything-small-hf
Prompt Depth Anything VITS Transparent	Depth Anything Team	depth-anything/prompt-depth-anything-vits-transparent-hf
LBM Depth	Jasper	jasperai/LBM_depth
DepthPro (Apple)	Apple	apple/DepthPro-hf
ZoeDepth (NYU+KITTI)	Intel	Intel/zoedepth-nyu-kitti
MiDaS 3.0 (DPT-Hybrid)	Intel	Intel/dpt-hybrid-midas
DPT Large (Intel)	Intel	Intel/dpt-large
DPT Large (Manojb)	Manojb	Manojb/dpt-large
DPT BEiT Large 512	Intel	Intel/dpt-beit-large-512
MiDaS v2 (Qualcomm)	Qualcomm	qualcomm/Midas-V2

Multimedia Framework

This project utilizes the FFmpeg multimedia framework for video/audio processing via subprocess invocation. FFmpeg is licensed under the GNU GPL v3 or LGPL, depending on how it was built. No modifications were made to the FFmpeg source or binaries — the software simply executes FFmpeg as an external process.

You may obtain a copy of the FFmpeg license at: https://www.gnu.org/licenses/

VisionDepth3D calls FFmpeg strictly for encoding, muxing, audio extraction, and frame rendering operations, in accordance with license requirements.