<h1 align="center"></h1> <p align="center"> <img width="450" height="263" alt="VisionDepth3D900x527" src="https://github.com/user-attachments/assets/eaae16ce-57ff-48f8-8dbd-1aca699b8724" /> <br> </p> <h2 align="center">The All-in-One 3D Suite for Creators</h2> <p align="center"> <em>This is Hybrid 3D.<br> Built from AI depth + custom stereo logic —<br> Designed for cinema in VR.</em> </p> <h3 align="center"> <a href="https://github.com/VisionDepth/VisionDepth3D/releases"> <img src="https://img.shields.io/github/downloads/VisionDepth/VisionDepth3D/total.svg" alt="GitHub Downloads"> </a> <img src="https://img.shields.io/badge/python-3.13-blue" alt="Python Version"> <img src="https://img.shields.io/github/last-commit/VisionDepth/VisionDepth3D" alt="Last Commit"> <img src="https://img.shields.io/github/stars/VisionDepth/VisionDepth3D?style=social" alt="GitHub Stars"> </h3> <p align="center"> <em style="font-size: 14px; color: #888;"> Click to download or support the project 💙 </em> <br> <a href="https://visiondepth3d.itch.io/visiondepth3d" target="_blank" rel="noopener"> <img src="assets/widget-preview.png" alt="Download VisionDepth3D on Itch.io" width="208" height="167" style="border-radius: 8px; margin-top: 6px;"> </a> <br> <p align="center"> <a href="https://visiondepth.github.io/VisionDepth3D/" target="_blank" rel="noopener" style="display:inline-block; margin-top:8px; padding:10px 16px; border-radius:12px; border:1px solid #39c6ff; background:linear-gradient(180deg, rgba(57,198,255,.15), rgba(57,198,255,.08)); color:#39c6ff; font-weight:700; text-decoration:none;"> Official website out now → </a> </p>

Notice

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

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

<h2 align="center">All-in-One 3D Suite</h2> <h3 align="center">3D Generator (Stereo Composer)</h3> <p align="center"> <img width="700" height="598" alt="3Dtab" src="https://github.com/user-attachments/assets/f6c82115-c3ba-464f-91c1-7a623ed11007" /> <br> <em>(3D Generator Tab)</em> </p>

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)

<p align="center"> <img width="700" height="598" alt="Depthtab" src="https://github.com/user-attachments/assets/c2b32af2-5be5-4d3f-9b41-cceb77b30785" /> <br> <em>(Depth Estimation Tab)</em> </p>

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.

FPS / Upscale Enhancer (RIFE + Real-ESRGAN)

<p align="center"> <img width="700" height="598" alt="frametools" src="https://github.com/user-attachments/assets/4abdc68f-b878-47b6-b185-2e39ace1ba1a" /> <br> <em>(FPS / Upscale Enhancer Tab)</em> </p>

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.

Depth Blender (Multi-Source Depth Fusion)

<p align="center"> <img width="700" height="598" alt="DepthBlendTab" src="https://github.com/user-attachments/assets/89c61a02-55e8-4ff6-8ed0-bd3bd739d04e" /> <br> <em>(Depth Blender Tab)</em> </p>

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.

Audio Tool (Rip, Attach, Attach + Stitch)

<p align="center"> <img width="558" height="587" alt="AudioTool" src="https://github.com/user-attachments/assets/bd7775a3-f625-4e77-be53-0f820a5f1b0b" /> <br> <em>(Audio Tool)</em> </p>

Rip: extract audio tracks from videos (copy or re-encode)
Attach: mux audio back into processed clips (fast copy by default)
Attach + Stitch: batch attach audio per clip, then stitch into one final gapless export
Smart matching: auto-match audio to clips by filename patterns or use one audio track for all
Audio offset control for sync fixes
Codec control for both per-clip muxing and final output encoding
Logging + progress via FFmpeg runner

Result: A practical post stage for restoring audio, correcting sync, and finishing multi-clip renders into one clean final movie.

Preview + Format Testing

<p align="center"> <img width="700" height="587" alt="3Dpreview" src="https://github.com/user-attachments/assets/4ce33583-8db7-40c2-a35d-d5c78efd26d9" /> <br> <em>(Live 3D Preview)</em> </p>

Real-time preview modes: Anaglyph, SBS, Passive Interlaced, Depth and Shift Heatmaps
On-frame tuning: convergence and parallax checks without committing to long renders
Save preview frames for quick comparisons and sharing

Smart GUI + Workflow

Multi-tab interface with persistent settings
Help menu + hotkeys
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

✔️ This program runs on python 3.13
✔️ This program has been tested on cuda 12.8
✔️ Conda (Optional, Recommended for Simplicity)

📌 Step 1: Download the VisionDepth3D Program

1️⃣ Download the VisionDepth3D zip file from the official download source. (green button)
2️⃣ Extract the zip file to your desired folder (e.g., c:\user\VisionDepth3D).
3️⃣ Download models Here and extract weights folder into VisionDepth3D Main Folder
4️⃣ Download Distill Any Depth onnx models here (if you want to use it) and put the Distill Any Depth Folder into Weights Folder

📌 Step 2: Create Env and Install Required Dependencies

🟢 Option 1: Install via pip (Standard CMD Method)

**1️. press (Win + R), type cmd, and hit Enter

VisionDepth3D

Install / Use

README