VisionDepth/VisionDepth3D

VisionDepth 3D – All‑in‑One 3D Suite for Creators

What it is – VisionDepth 3D is a Windows‑only desktop application that turns 2‑D photos or videos into stereoscopic 3‑D (side‑by‑side, anaglyph, VR180, etc.). It does the heavy lifting locally with AI‑driven depth estimation, depth‑map blending, frame‑interpolation (RIFE) and AI super‑resolution (Real‑ESRGAN/BSRGAN). The UI is split into tabs that guide you through each stage of the pipeline.

Core components

Component	What it does	AI/ML part
Depth Estimation	Generates per‑pixel depth maps for images, videos or frame folders.	Supports many open‑source depth models (Depth Anything v1/v2, Distill‑Any‑Depth, Video Depth Anything, Marigold, ZoeDepth, DPT, etc.) via PyTorch, ONNX or Diffusers back‑ends.
Depth Blender	Fuses two depth sources into a cleaner map.	Simple weighted blending plus edge‑preserving denoise and CLAHE contrast shaping.
3D Generator / Stereo Composer	Projects the depth map into a stereoscopic view using either a classic per‑pixel warp or the newer Subpixel Z‑Splat renderer.	Uses the depth map to compute per‑pixel parallax, with controls for pop‑out, convergence, edge masking, and occlusion healing.
FPS / Upscale Enhancer	Interpolates frames (RIFE) and upsamples the result (Real‑ESRGAN/BSRGAN).	RIFE 4.25 (Lite & full) for 2×/4×/8× frame‑rate boost; AI super‑resolution models auto‑selected based on output size.
Live 3D	Real‑time preview of depth‑based stereo from a webcam, capture card or screen capture.	Same depth models as above run on‑the‑fly; useful for testing before committing to a full render.

Free vs Pro

• Free tier – 3‑minute video limit, 1080p max height, watermarked output, basic depth models, 2× RIFE, no batch processing.
• Pro tier – One‑time $59.99 unlock; removes limits, adds batch/queue workflows, 4K+ output, full model list, high‑FPS RIFE, AI super‑resolution, VR/VR180 export, advanced stereo controls, job queue, and no watermark.

Installation

1. Download the VisionDepth 3D Setup Hub from the project's Itch.io page.
2. Choose the appropriate build (CUDA for NVIDIA GPUs – recommended, or DirectML for AMD/Intel GPUs).
3. The hub downloads a signed installer (.exe) and required binary blobs, then runs the Inno Setup wizard.
4. After installation, launch VisionDepth 3D from the desktop shortcut.

Typical workflow

1. Import an image or video.
2. Generate depth – pick a model, let the app cache it locally, preview the depth map.
3. (Optional) Blend a second depth source to improve quality.
4. Configure stereo – adjust strength, pop‑out, convergence, etc., and preview side‑by‑side or anaglyph.
5. Enhance – run RIFE for smoother motion and/or AI super‑resolution for higher spatial detail.
6. Export – choose container (MP4/MKV/AVI), stereo format, resolution, and let FFmpeg encode (CPU or hardware‑accelerated). The app can also queue multiple jobs.

Who it’s for

• Independent filmmakers and VR creators who need a quick, offline way to convert legacy 2‑D footage into 3‑D.
• Content creators making YouTube‑style 3‑D videos, anaglyphs, or VR180 experiences.
• Hobbyists experimenting with depth‑aware visual effects without writing code.

Licensing

• The application is proprietary, distributed under a “no‑derivatives” license; the source code is not public.
• All AI depth models are third‑party and remain under their original open‑source licenses (Apache‑2.0, MIT, etc.). Users must comply with those licenses separately.

Key take‑aways

• VisionDepth 3D bundles a full 2‑D‑to‑3‑D pipeline (depth, blending, stereo rendering, interpolation, upscaling) into a single GUI.
• It runs locally, so no media leaves the user’s machine.
• The free tier is functional for short, low‑resolution experiments; the paid tier unlocks professional‑grade features and removes output restrictions.