Changelog
- Added PeakMagnetar, OptiMagnetar, and TinyMagnetar, three efficient AI models for high-quality image and video upscaling. They support 1×, 2×, 4×, 6×, and 8× for images and 2× and 4× for video. Image upscaling preserves transparency and color profiles.
- Added target resolutions for video upscaling. Choose a fixed output size when you need exact dimensions. If the aspect ratios differ, Visual Enhancer adds black bars to keep the whole picture visible. Export controls disable codecs that can’t handle the selected dimensions, preventing encoder failures on large 4× exports.
- Improved high-resolution video playback in Compare. On supported clips, decoded frames stay on the GPU through display, reducing transfers through system memory. Compare retains the original source resolution for inspecting fine detail. This playback path supports compatible NVIDIA, AMD, and Intel GPUs.
- Added VP9 and WebM export. VP9 defaults to WebM and can also be exported to MKV or MP4. WebM supports AV1, preserves compatible audio, and converts supported text subtitles to WebVTT.
- Improved image metadata handling. Supported image workflows preserve EXIF and XMP more reliably, including when filenames contain spaces or non-Latin characters.
- Improved automatic export names and batch feedback. Auto naming produces
photo_visual_enhancer.pngand adds_2,_3, and so on when a name is already taken. The processing speed display now rounds to the nearest tenth of an FPS. - Expanded interface translations across all 11 supported non-English languages to cover the new model controls, export options, and status messages.
SharpNext has been discontinued and is no longer available in the Upscale model selector. For image upscaling, choose VisionNext, RazorNext, or one of the three new Magnetar models.