Skip to content

DirectML - VisionDepth3D v5.2 Official Release

Choose a tag to compare

View all tags
@VisionDepth VisionDepth released this 25 Sep 18:14
DirectML-v5.2.0
995b818
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

VisionDepth3D v5.2

Better Live 3D. Improved depth blending. A workspace you can make your own.

VisionDepth3D v5.2 brings improvements to rendering quality, usability, and compatibility across the CUDA and DirectML builds.

What’s New

Live 3D Overhaul

Live 3D is more responsive and stable, with reduced waviness, improved scene-change handling, and fixes for reversed depth and playback crashes.

The familiar 3D Assistant now comes to Live 3D, including named styles, six assistant controls, and adjustments during playback. Switch styles while watching or transfer your current settings from the main 3D tab.

Live 3D requires Pro. Performance depends on your hardware and selected settings.

Improved Depth Blending

The new Edge-Aware blending mode preserves the base depth map while selectively adding detail from a second map.

New depth-range alignment and edge-protection controls give you more control over the result. A Reset Settings button restores blend defaults without clearing your selected files.

Legacy blending remains available.

Customizable Workspaces

Arrange your workspace with movable, collapsible, and floating panels.

Show or hide panels, save your layout, and restore the default arrangement with Reset Layout.

Cleaner Z-Splat Rendering

Further refinements reduce ghost contours, halos, and flashing around moving edges.

Subpixel Z-Splat remains CUDA-only. DirectML continues to use Classic Stereo Warp.

HDR10, FFmpeg, and Workflow Fixes

Improved HDR10 processing and encoding compatibility, updated frame extraction for newer FFmpeg builds, and corrected extraction progress reporting.

HDR10 rendering with NVENC was successfully tested during development. Compatible encoding hardware and drivers are required.

Interface and Language Polish

Expanded translations, improved language switching, refreshed startup artwork, and updated in-app release information round out the release.

Choose Your Build

  • CUDA: Recommended for supported NVIDIA GPUs.
  • DirectML: Intended for supported AMD and Intel GPUs.

Both editions receive the shared v5.2 workflow and interface improvements. Feature support and performance vary by backend.

Installing v5.2

  1. Open the VisionDepth3D Setup Hub.
  2. Uninstall your currently installed VD3D build.
  3. Download the v5.2 CUDA or DirectML installer.
  4. Keep the installer and any required .bin files together.
  5. Run the installer and complete setup.

Use installer builds from the official VisionDepth3D release channels.

The existing Free and Pro structure remains unchanged. Upscaling and Live 3D require Pro.

Full Changelog

For the complete list of changes, fixes, compatibility notes, and testing details:

[Read the full v5.2 changelog](https://github.com/VisionDepth/VisionDepth3D/blob/Main-Stable/CHANGELOG.md)

Thank You

Thank you to everyone testing builds, sharing renders, reporting issues, and supporting development through Pro.

If you encounter a problem, include your build, GPU, selected settings, and relevant logs when reporting it.

Enjoy creating with VisionDepth3D v5.2!

Assets 4
Loading