Changelog
- Neuroframe Engine (Upscale) performance gain:
- 720p60 → 1440p Ultra: 10.7-11.9 FPS → 113.38 FPS (9.53-10.60× faster).
- 1080p30 → 4K Ultra: 4.2 FPS → 52.07 FPS (12.40× faster).
- Neuroframe Engine (Frame Interpolation) performance gain:
- 720p 30→60: 123.08 FPS → 247.65 FPS (2.01× faster).
- 1080p 30→60: 74.86 FPS → 131.52 FPS (1.76× faster).
- 4K 30→60: 12.08 FPS → 36.93 FPS (3.06× faster).
- Completely reworked the Neuroframe Engine video Upscale pipeline, delivering a massive increase in processing performance by keeping much more of the video processing directly on the GPU.
- Completely reworked Frame Interpolation processing with a new DLSS Frame Generation pipeline designed for significantly higher throughput, with up to a 3.06× performance increase at 4K.
- Added Live mode, allowing videos to be watched with DLSS 5 Neural Rendering while supported parameters can be changed and applied in real time during playback.
- Added automatic Processing Engine Path selection, replacing the old manual VRAM / RAM option. v9.0 now automatically selects the best available processing path based on the selected codec and workflow.
- Added automatic video container selection, automatically choosing MP4, MKV, or MOV depending on the selected video codec.
- Significantly improved multi-GPU video processing, allowing the processing GPU and encoding GPU to work together more efficiently when different GPUs are selected.
- Improved Frame Interpolation source detection, including better handling of variable frame rate videos for more accurate frame rate detection and output generation.
- Improved Frame Interpolation output verification, automatically checking the final frame count, frame rate, resolution, HDR, and 10-bit output before processing is marked as successfully completed.
- Improved HDR and 10-bit video processing, with additional validation to make sure high bit-depth and HDR output is preserved correctly.
- Improved Image Upscale batch processing by reusing the Neuroframe Engine between compatible images, reducing unnecessary initialization and improving processing efficiency across multiple files.
- Improved Neural Rendering real-time previews, with previews now automatically refreshing when Neural Rendering settings are reset.
- Changed Frame Interpolation to use NVIDIA NVENC encoding by default, providing faster GPU-accelerated video encoding on supported NVIDIA GPUs.
- Improved overall video processing reliability, including stronger checks for decoded frames, output files, frame timing, encoding, and processing completion.
- Improved preset and settings compatibility for v9.0, including automatic handling of older settings that are now controlled automatically by the Neuroframe Engine.