DLSS 5 Visual Enhancer v6.0

Changelog

• Added Live mode for watching local videos, direct streams, YouTube, and Twitch with DLSS 5 Neural Rendering applied during playback. Neural Rendering parameters can be changed in real time and applied instantly without restarting the session.
• Up to 163% speedup for Frame Interpolation processing across all supported GPUs.
• Up to 70% speedup across all processing types on Ada Lovelace and Blackwell GPUs with new architecture-optimized DLSS 5 runtimes.
• Completely reworked batch processing and outputs. Process individual files or entire folders directly from disk, choose your own output folder, and handle large batches with much less unnecessary browser preview and download overhead.
• Added a new DLSS 5 Architecture setting with optimized runtimes for Turing/Ampere, Ada Lovelace, and Blackwell GPUs. Auto mode automatically selects the best version for the detected GPU.
• Reworked Frame Interpolation processing so decoding, motion preparation, DLSS Frame Generation, and encoding can work together more efficiently instead of waiting for each stage to finish one by one.
• Expanded Frame Interpolation up to 480 FPS, with new 119.88, 144, 165, 180, 240, 360, and 480 FPS output options.
• Added detailed per-file batch progress with Queued, Running, Complete, Failed, Cancelled, and Skipped states, together with progress percentage, elapsed time, output location, and processing details.
• Improved batch cancellation. Completed files are preserved, interrupted outputs are safely cleaned up, and remaining files are clearly marked as skipped.
• Added more control over Live playback, including source quality, processing resolution up to 4K, automatic or fixed frame rate, playback buffering, and Fast or Quality motion guide modes.
• Added automatic Live performance monitoring with processing speed, playback buffering, dropped frames, rebuffer events, and audio/video synchronization information.
• Added DLSS 5 render settings to output metadata for supported image and video formats, making it easier to identify which Neural Rendering settings were used later without replacing existing user metadata.
• Improved video processing for files with missing or incorrect frame-count information, reducing failed or incomplete renders caused by unreliable video metadata.
• Improved final output verification so rendered videos are checked before being marked as complete, helping prevent incomplete or corrupted outputs from being presented as successful.
• Improved output safety across Image, Video, and Frame Interpolation processing. Finished files are only published after processing and verification complete successfully.
• Improved Stop and cancellation behavior across rendering, previews, video probing, encoding, and final output processing for faster and cleaner interruption of active jobs.
• Added automatic cleanup of old temporary files, caches, abandoned render jobs, and Live session data, reducing leftover files from previous or interrupted sessions.