This is a medium to low effort filtering and encoding for Amnesia. slow.pics compare: S01 https://slow.pics/c/zStsUbpl OVA https://slow.pics/c/ihJ8GnB4
For the main season, the quality of the BD is fairly good. We first vsdenoise.frequency_merge the OP of all episodes together using a filtered arithmetic mean for low frequencies to get rid of artefacts caused by starving in the source, and a power mean for high frequencies. After the merge, we uses bore to remove border ringing.
For the main season, we then perform a faux descale to deal with the halo and the aliasing in the source. The lineart is rough even after the descale, and there are also some aliasing in some of the assets used, but we for now decide to keep it as is and don't perform further AA. The descale is protected by a complex custom expr to clamp the result to the source, and a credit mask based on PaddleOCR based credit mask.
On the other hand, for the OVA, there is only a DVD available. The DVD has heavy ringing from kernels. To get rid of the ringing, we use lvsfunc.LHzDelowpass.DoubleTaps_4_4_125_1375_mpeg2 to delowpass. We use a variance mask to prevent the model from destroying texture.
The banding in this anime can be quite severe occassionally. For both the main season and the OVA, to handle the banding as much as we can while minimising damage otherwise, we calculate CAMBI score for each frame, and apply different strength of deband via vszipcu.Deband based on the CAMBI score calculated. We then use MVTools to stabilise the video including the debanding result before sending the video to encoder.
For encoding, we use 5fish/SVT-AV1 v2.3.260819 for the encode. Targeting a higher quality, we use --crf 16.00 for the main season and --crf 14.00 for the OVA, with --lineart-psy-bias 7 --texture-psy-bias 4 as the base tuning. SVT-AV1 is still slightly prone to creating mosquito noise sometimes with this source. For the main season, because high DLF strategy tends to destroy details in very dark scenes, and this anime has a lot of them, we instead rely on CDEF and use --cdef-bias-max-cdef 4,2,2,0 --texture-cdef-bias-max-cdef 2,0,0,0 to mostly achieve a clean encode with minimum loss. For the OVA, we use --satd-bias 0.50 instead. At last, we apply a finer layer of grain to balance everything.
Thanks to PingWer74, Artorius, Zastin, bouquet, Zahuczky, jensen2267, and other people from the JET server for teaching me the way to handle issues with this DVD. Thanks to all the people creating the tools, plugins and vsfunc's that make this possible, and thanks to everyone in the community for teaching me, pointing out issues, and giving me advice.
If you're interested in AV1 encoding for anime, you are welcome to join AV1 weeb edition at https://discord.gg/83dRFDFDp7.
You are welcome to do anything with this encode, including remuxing and republishing it. However, we would ask you to leave a credit to Kekkan if you can.