aaronfeng
patreon
v3.116.13-beta [Premium] Huge performance boost for Real-ESRGAN
🕑 Added 2024-08-17 14:19:30 +0000 UTCThis is the Premium version, Premium features are Unlocked, you can use it Commercially, and will NOT ask you for donation.
This a BETA version, click here to get Stable version.
You need to active the premium version for your PC, you can read this post to learn more: How to activate the Premium version
"Why my download got flagged as unsafe?". Click here to learn why there is "False positive antivirus alerts"
*You DO NOT need to re-activate the software if you already activated it, if you uninstall then re-installed the software, just enter the old activation code.
-
FULL Package:
-------------------------------------------------
Update Package (For v3.116.01 and Later versions):
You must fully installed v3.116.01 and Later versions before install the Update Package.
-
Sha256 (FULL Package):
29baab3a7106f0d4f4ec2ad19eb749968c8736142136217676713f4e738e4a49
Sha256 (Update Package):
480dcaeda9ed8ffda1010eb053dfe30b18f653c3e72f4593e03d12e78e1dfe6c
-
v3.116.13-beta
- 🚀Upgrade Real-ESRGAN [Sequential decoding] to [Image decode threads]. Boost image upscaling performance by auto adjust decode threads based on PCs hardware, up to 87.94% faster during testing.
In my tests (involving 2x upscaling multiple folders of images), this new feature has boosted image upscaling speed by up to 87.94%. Results may vary depending your PCs hardware, the model & upscale ratio you choose and your images.
- ✅Fix bug: In some situations, number of frames does not meet expectation after frame interpolation.
Comments
Aaron Feng
More video encoding options will be available in the future
Aaron Feng
Yes, and thank you for your suggestion
Aether
when possible mind adding EXR/HDR formats? with 32bit color support
Gesandro
Cool that its now available for testing. Would be great if you could adapt this new procedure to the video "enhancing" too. And to spread some knowledge and correct your calculations. The new method is not 88 % faster, the older is 88 % slower. Or the other way around the new method is 46.79 % faster. Its importend to use the correct context, because otherwiser we would have by assumed 132s run time of the new method a 100 % faster method. If you think about it it would be crazy because that means no calculation time but we had some calculation time... aaaahh math. Sorry for beeing a smartass, good work keep it going :)
