VisionDepth3D v3.1.6 – Release

Performance gains, smarter resizing, and enhanced pipeline reliability in this focused update!

What’s New?

Core Logic & Backend

• FFmpeg Backend Added for upscale processing in the frame renderer
  • Up to 4–6× FPS boost in render speed
  • Fallback logic when GPU-based ESRGAN is disabled
• Input Resize Correction
  • Auto-resizes input to match GUI render res
  • Fixes cropping/stretching mismatches across RIFE and Upscale pipelines
• GPU Interpolation Integration
  • Full TensorRT support in Upscale pipeline
  • VRAM-safe memory handling & batch-aware resolution scaling
• NVENC Export Polishing
  • Improved export reliability with hardware encoders
  • Graceful handling of odd resolutions (e.g., 1280×719 → safe round)

GUI & Preview Enhancements

• Updated Preview Layout (WIP) – First pass at organizing layout for clarity
• Depth ETA Fix – ETA now shown in HH:MM:SS format
• Elapsed Time Tracker – Tracks full batch/video runtime

Why This Matters

• Massive upscaling performance boost
• Better visual feedback during long renders
• Fixes common bugs related to frame sizing

Full Change Log

• Added ffmpeg-based upscaling logic for renderer
• Fixed input/GUI resize mismatches
• Integrated ONNX RIFE interpolation into render/export
• Improved NVENC support for edge-case resolutions
• ETA logic now shows hours for long renders
• Elapsed render time added to GUI
• WIP: Preview panel layout revamp
• Experimental divergence tracking scaffolding

💙 Thanks to early testers and notebook devs for profiling, feedback, and pushing this forward!

This update sets the stage for full pipeline modularity in the next major version.