Download for Windows.
Get the complete three-file installer set. Keep every file in the same folder, then run the small Setup EXE.
Before you download
- 64-bit Windows 10 or Windows 11
- CUDA-capable NVIDIA GPU with a current driver
- About 6 GB for the installed app, plus downloads and working space
- 16 GB RAM and 6 GB VRAM recommended
This build does not run on 32-bit Windows, ARM Windows, AMD-only graphics, or Intel-only graphics.
Download all 3 files
StereoLive3D-0.11.2-NVIDIA-Beta-Setup.exe2.68 MiB · SHA-256: 910cf09dda558add75cc825539bb6a1eada542a1e39686803e81df21f55c2498StereoLive3D-0.11.2-NVIDIA-Beta-Setup-1.bin1.86 GiB · SHA-256: 93d57bbcef32ab98fa984ad8ada465092e285b962ada013c9e18dc109838a09bStereoLive3D-0.11.2-NVIDIA-Beta-Setup-2.bin737.50 MiB · SHA-256: 08f421e2abe19a0f086636466f9bda9664ecaa3ee5fa36195002fcf39fbb7307Install in three steps
- 1
Put the EXE and both BIN files together in one folder. Do not rename them.
- 2
Run the Setup EXE. Because this beta is not code-signed, Windows may show a SmartScreen warning.
- 3
Open StereoLive 3D, choose Balanced, select the displays, and start your trial session.
LiveKit browser streaming
- Bundled media server and NVIDIA NVENC video with Opus audio
- Selectable playback and recording audio, plus video-only fallback
- GUI bitrate maximum and optional adaptive bitrate (30 Mbps starting default)
- Manual fullscreen SBS controls in the browser player
- Existing Virtual Desktop/window and legacy HTTP/HLS modes retained
Python and separate FFmpeg are not required for LiveKit. Browser playback and audio sync remain beta features; LiveKit does not forward desktop controller input.
Follow the Quest setup tutorialEnable NVIDIA TensorRT
StereoLive 3D works with its bundled NVIDIA CUDA backend. Installing TensorRT 10 can increase frame rate by accelerating depth inference.
- 1
Download the Windows ZIP package for NVIDIA TensorRT 10 from the official NVIDIA installation guide.
- 2
Extract the package. Copy the contents of its
binfolder intoC:\TensorRT\bin. Keep every DLL filename unchanged. - 3
Restart StereoLive 3D and leave Processing backend set to Automatic (TensorRT / NVIDIA CUDA).
- 4
Start a session and confirm that the log says
Selected depth backend: TensorRT.
If the log says NVIDIA CUDA, TensorRT was not detected or could not initialize. StereoLive 3D will continue using its compatible CUDA fallback.
The first TensorRT session may take longer while it prepares and caches an optimized engine for the selected depth resolution. Later sessions reuse that cache. Advanced users may keep TensorRT elsewhere and set TENSORRT_PATH to its bin directory.