Changelog
- Added a completely new command-line interface with VE_CLI, allowing Neural Rendering, Upscale, and Frame Interpolation jobs to be run directly from the terminal with batch inputs, presets, GPU selection, output controls, dry runs, quiet mode, JSON results, and GPU device listing.
- Added full interface localization with English, Arabic, German, Spanish, French, Hindi, Japanese, Korean, Polish, Ukrainian, Simplified Chinese, and Traditional Chinese, including proper right-to-left interface handling for Arabic.
- Added a new Rolling Cache (5 GB) processing mode alongside the existing Stage by Stage workflow, allowing multi-stage video processing to continuously pass frames between stages without creating a complete intermediate video after every stage. Cache intermediates can now use either FFV1 or ProRes Proxy.
- Completely redesigned the Batch Queue with a clearer table-based layout, larger media thumbnails, richer file and processing information, per-file progress, processing FPS, estimated remaining time, completed processing time, improved status reporting, automatic selection tracking, and a new option to retry failed or cancelled jobs.
- Significantly improved video preview and playback with a new high-precision preview decoding path designed to preserve 10-bit and HDR video information, more accurate frame-based seeking, previous/next-frame navigation, a redesigned timeline ruler, rendered preview ranges shown directly on the timeline, improved synchronization between input and processed footage, and integrated volume controls.
- Expanded automatic Color Adjustment for video with separate Auto 1 Frame and Auto Fixed Frames modes, allowing adjustments to be calculated from the current playhead frame or from multiple fixed samples across the video for a more representative overall grade.
- Added a new Reference Image matching workflow for Color Adjustment, automatically analyzing the source and a selected reference image and fitting the editable exposure, contrast, tone, white balance, hue, vibrance, and saturation controls to reproduce the reference color distribution while keeping the result fully adjustable and exportable as a LUT.
- Reworked RTX Video HDR inside Neural Rendering into its own independent processing stage instead of coupling it to RTX Super Resolution. RTX Video HDR can now be enabled, disabled, and reordered separately within the video processing pipeline while preserving existing v12 HDR settings during migration.
- Improved GPU and video processing device selection with a dedicated Decoding / Encoding Device setting, providing more flexible CPU, FFmpeg, Vulkan, and hardware-accelerated processing selection independently from the GPU used for Tensor Core / NGX AI processing.
- Improved portable application behavior, keeping application-owned temporary files, caches, profiles, and runtime data within the portable installation and making stored paths more resilient when the application folder is moved to another location.
- Removed the standalone Denoising stage from Neural Rendering. Existing saved settings and workflows are automatically migrated to discard the retired stage, while NVIDIA NIS and AMD CAS remain available as sharpening methods.
- Improved overall performance, stability, and workflow reliability across FFmpeg processing, preview caching, video filters, sharpening, export validation, window positioning and focus behavior, queue management, GPU handling, processing progress, and general interface responsiveness.