Based on machine learning, LS1 is a small and fast general-purpose neural network optimized for scaling factors in the 1x–2x range. LS1 has similar performance and image quality to FSR, while surpassing it in many cases.
There have also been many other changes since the release of version 2.0, the last major announcement. Summary for those who do not follow updates in the community:
- New scaling types: Anime4K, Sharp bilinear, xBR
- Cursor scaling option
- Latency improvements
- Experimental VRR support
- Stretch to fullscreen mode
- Windowed mode
- Crop input option
- Multi-display mode
I look forward to your feedback,
Have fun!
More from Lossless Scaling
Other announcements
Performance Mode HotfixResolved a critical bug that impacted quality in LSFG 3.1 Performance mode. · Fixed an issue where holding down a key wouldn’t repeat input if it was part of a scaling hotkey combo. · Added Macedonian localization. Big thanks to everyone helping make LS better. Also, check out the excellent work by PancakeTAS on running LSFG on Linux: Have fun!LSFG 3.1This update introduces significant architectural improvements, with a focus on image quality and performance gains. Quality Improvements · Enhanced overall image quality within a specific timestamp range, with the most noticeable impact in Adaptive Mode and high-multiplier Fixed Mode · Improved quality at lower flow scales · Reduced ghosting of moving objects · Reduced object flickering · Impro…Adaptive Frame GenerationAFG Introducing Adaptive Frame Generation (AFG) mode, which dynamically adjusts fractional multipliers to maintain a specified framerate, independent of the base game framerate. This results in smoother frame pacing than fixed multiplier mode, ensuring a consistently fluid gaming experience. AFG is particularly beneficial for games that are hard or soft capped at framerates that don’t align as…