Hello VD3D Users!

It has been a busy cycle behind the scenes, and I am excited to finally say that VisionDepth3D v3.7 is out now.

The last versions each pushed VD3D in different directions:

• v3.5 focused on cinematic polish, DOF, audio tools and the overall look of the output.
• v3.6 brought depth blending, HDR10 preservation, and a much faster Frames pipeline.

With v3.7, the focus is stability, cross platform support and real time 3D. Live 3D Capture has been overhauled, the Depth tab has a new unified backend and codec system, and the 3D and Depth Blender workflows are cleaner and easier to tune.

Here is what is new.

Live 3D Capture and Floating Window Refinement

Live 3D Capture

Live 3D Capture has been given a full pass for reliability and usability.

• Optional live audio passthrough monitor for external capture devices, with device selection and delay control.
• Audio routed through DirectShow and WASAPI, with an FFplay based monitor for low latency listening.
• New color channel controls fix purple / red tint issues on some capture cards.
• Tuned depth inference defaults for more practical FPS on 1080p HDMI sources.
• New --no-preview headless mode to remove local display overhead and prepare for streaming workflows.
• Architecture prepared for future browser based SBS VR streaming with synced audio and video.

Floating Window and Temporal Stability

The stereo presentation has been tightened so depth feels stable instead of “breathing.”

• Rebuilt Dynamic Floating Window so it masks only one edge, based on parallax direction.
• Adds a minimum parallax threshold so the window stays off when depth is close to the screen.
• Uses temporal smoothing and easing so the window glides in and out instead of popping.
• Fixes jitter and “in and out” depth pulsing by smoothing convergence and subject depth over time.

Result: fewer edge violations, less shimmer and a more comfortable stereo experience in both VR and flat playback.

Black Bar Detection

• Improved auto crop logic for letterboxed content.
• Handles fade ins, fade outs and darker scenes without false detections.
• No more vertical drift when black bar height changes.

Unified Depth Pipeline and Platform Support

The Depth tab now runs on more hardware and gives you more control over how depth videos are encoded.

Multi Backend Depth Inference

Device detection has been rewritten so CUDA is no longer assumed.

Supported backends now include:

• CUDA on NVIDIA GPUs
• ROCm on AMD GPUs
• MPS on Apple Silicon
• CPU fallback for systems without a GPU

This reduces crashes on AMD and macOS, avoids accidental CPU only runs on powerful GPUs, and lays the groundwork for future Linux builds.

Codec Selection and Control

• New Video Codec dropdown for depth video exports.
• Choose from hardware encoders (NVENC, AMF, QSV) or CPU encoders (libx264, libx265, AV1, and legacy MPEG-4 variants).
• Safer behavior for XVID and other codecs on non NVIDIA systems.
• Guard rails and warnings where AV1 decoding is limited.
• Codec support is now aligned across Depth, 3D Generator, and FrameTools.

Depth Job Control

• Depth renders now support Pause, Resume and Cancel.
• Pauses release resources more gracefully and cancels avoid corrupt output.
• Clear status messages for running, paused, canceling and completed states.

3D Pipeline and UX Improvements

The main 3D pipeline is more flexible and easier to understand.

• New Keep Original Audio option to copy the source video’s audio into the final 3D render without re encoding.
• New image based 3D pipeline that runs through the main renderer, ideal for single frame stereo stills.
• The Mode selector now cleanly switches between Single, Batch and Image workflows.
• A new 3D filename suffix system automatically labels exports by format and eye mode, for example:
  _LRF_Full_SBS, _LRF_Half_SBS, _VR, _Anaglyph, _Interlaced, _LRF_Left, _LRF_Right.
• Labels and tooltips across multiple languages have been reviewed and cleaned up.

Depth Blender Preview and Scrubber

The Depth Blender tab is now a visual tool instead of a blind tweak and pray step.

• Live preview shows the V2 base map and the blended result side by side.
• All blend parameters (white strength, feather blur, CLAHE, bilateral filters) update the preview in real time.
• A frame scrubber lets you move through a sequence and see how the blend behaves across time before running a full batch.

Closing Thoughts

This update is about three main pillars:

1. Stability
   Live 3D Capture, floating window behavior and temporal depth stability have all been tightened so 3D feels more consistent and comfortable.

2. Platform reach and control
   The Depth tab now supports NVIDIA, AMD ROCm, Apple Silicon and CPU only setups with flexible codec selection and job controls.

3. Better preview and workflow
   The 3D pipeline, audio handling and Depth Blender preview all make it easier to test, tune and export your 3D content with confidence.

I will continue to refine the Live 3D path, streaming options and presets in future builds. Feedback on the new capture behavior and the cross platform depth pipeline is very welcome.

Also, VisionDepth3D has now passed over 3,000 downloads, which is wild to think about. Thank you to everyone who has been testing, reporting bugs, and helping shape this tool.

How to Install

1. Go to the VisionDepth3D Releases page
2. Download the latest v3.7 installer .exe and .bin parts
3. Place all files in the same folder
4. Run the .exe installer and follow the prompts
5. Launch VisionDepth3D from the Start Menu or desktop shortcut

For source installation and advanced setup, see the Installation Guide.

Have a feature request or found a bug
Please open a thread in the GitHub Discussions so we can track it and keep improving VisionDepth3D together.