OK, hmm screenshot of the mask? As well maybe I could take more clearer pictures but I don’t remember the settings to do such in my camera
Hmm. It’s something in grade causing this. The only 4 new settings that I’ve added in this time around within the grade shader are Black Level, Black-Red Tint, Black-Green Tint and Black Blue Tint. Maybe try defaulting those 1 by 1 to see if anything happens. I’m trying to diagnose the issue first if it’s actually a setting causing the issue like the contrast setting did previously.
Try defaulting all 4 of those. If it’s those 4 I’ll put out another pack right away and get rid of those settings.
A normal screenshot (F8 key) would be OK, it displays minute details on what is going on on the screen.
this is a base problem at original preset level, i think it might not have nothing to do with your prreset in this case sonkun, it might be my screen res or something
New Release Version (2024-09-15-r1):
Notable changes:
- NTSC shaders optimizations, cleanup.
- Improvements on NTSC coloring/rainbowing implementation.
- Two new NTSC coloring/rainbowing modes.
- Low-tap count behavior improved (NTSC).
- Internal resolution option added to NTSC shaders, similar to the HD feature.
- NTSC Sharpen changes to the ‘positive’ mode, now both modes dissolve dithering patterns, positive mode preserves more details. Use resolution scaling feature if missing the ‘old positive mode’.
- Pre-scaler presets included in the ‘guest’ folder. They are optimized for prepending to crt presets. Usually regular scaler presets scale to viewport, which can be very slow and also excessive.
- Most existing presets shouldn’t be affected (@sonkun), some testing is welcome anyway.
Download link:
https://mega.nz/file/tk4xFBCL#-3qo6saKm86uIAT9uF4jipKzvEX1uL3aOSSQVT6EWns
Example of tuned NTSC aperture preset, using new features:
- horizontal pre-scaler
- 8-taps
- rainbowing
- positive NTSC sharpen
it looks good. But what am I saying… every update is good!
Nice. I knew that last update wasn’t going to be the final one. So right off the bat loading this update up with my go-to test game Sonic 1 I notice 2 things, first was that I noticed a speed increase, no more jankiness and dropping of frames, it’s full speed again using my rainbow presets. The second thing I noticed loading up one of my composite presets were these vertical lines in the water that wasn’t there previously:
I seemed to track it down to the “ntsc gamma correct” setting. It’s more severe after turning it up to 0.12:
After disabling it they dissapear:
I see the new rainbow options as well. Off first glance setting 1 seems to be the new iteration that you introduced in the previous update, setting 2 seems to be the older original iteration you first introduced and setting 3 seems to be a slightly different iteration of setting 1. Not sure which one I’ll end up sticking with at the moment but it’s nice to have more options to choose from.
Any chances to incorporate Noise options from PlainOldPants, please ?
Thanks for testing, seems like a regression indeed. I will give it my best later this day…
Edit:can you test this version?
https://mega.nz/file/Y5pCGJTT#wYFExF6r2V6eZ0W5P7CEvZzwXNTCrxt-d2P-sW5Vxm0
I’ll try it right now give me a sec
Edit: @guest.r Well that was a quick fix, no more vertical lines in the water:
New Release Version (2024-09-16-r1):
Notable changes:
- NTSC shaders optimizations, cleanup.
- Improvements on NTSC coloring/rainbowing implementation.
- Two new NTSC coloring/rainbowing modes.
- Low-tap count behavior improved (NTSC).
- Internal resolution option added to NTSC shaders, similar to the HD feature.
- NTSC Sharpen changes to the ‘positive’ mode, now both modes dissolve dithering patterns, positive mode preserves more details. Use resolution scaling feature if missing the ‘old positive mode’.
- Pre-scaler presets included in the ‘guest’ folder. They are optimized for prepending to crt presets. Usually regular scaler presets scale to viewport, which can be very slow and also excessive.
- Edit: luma fix with gamma corrections and adaptive sharpen (thanks @sonkun for testings).
- Edit2: clamping bug fix
Download link:
https://mega.nz/file/wtJ1wYaa#9aSsJVQdgw_gxZiZ7eI6Qt8aFqQ9JMCt-WUuhZ4irnQ
No problem, glad I was able to help and thank you for the fast new release. I also smell a new sonkun pack in the making to go along with this update. Gonna take this to the lab and see if I can squeeze out a little more sharpness in my composite presets and further check out the new rainbow options.
Ok so upon trying out the new “positive” ntsc sharpness at a value of 10.00 with Sonic I noticed something strange, whenever I move Sonic forward or back and then stop a very small amount of pixels seem to “glitch” for just a few seconds. I only notice it in this small white area between his nose and eyes:
I tried to capture video footage of it:
This also happens if I load up the standard guest ntsc shader and change ntsc sharpness to 10.00 as well. That also doesn’t happen at all at -10.00.
Thanks for a keen eye and more testing. It seems it’s a precision error involved, but looks like it’s getting fixed. Found some more stuff to fix, so it’s getting there…
Glad I could help. When you’re in the lab for hours at a time I guess even these small kind of things become noticeable and sticks out like a sore thumb when you’re constantly in and out of the parameter settings lol. I’ll be around to test any test packs you may have.
New Release Version (2024-09-16-r2):
Notable changes:
- NTSC shaders optimizations, cleanup.
- Improvements on NTSC coloring/rainbowing implementation.
- Two new NTSC coloring/rainbowing modes.
- Low-tap count behavior improved (NTSC).
- Internal resolution option added to NTSC shaders, similar to the HD feature.
- NTSC Sharpen changes to the ‘positive’ mode, now both modes dissolve dithering patterns, positive mode preserves more details. Use resolution scaling feature if missing the ‘old positive mode’.
- Pre-scaler presets included in the ‘guest’ folder. They are optimized for prepending to crt presets. Usually regular scaler presets scale to viewport, which can be very slow and also excessive.
- Edit: luma fix with gamma corrections and adaptive sharpen (thanks @sonkun for testings).
- Edit2: clamping bug fix.
- Edit r2: ‘glowing nose’ problem mitigation/fix, some code improvements.
Download link:
https://mega.nz/file/YwoGyZjY#5CnIna_bBCob553IvmhQtdoNu0hXSi5vg4cfipdSVys
Yup, tested it right away and that problem is gone. Ok I’m gonna continue playing around and report back if I find anything.
Edit: @guest.r On another note I’m noticing the difference between negative ntsc sharpness and the new positive ntsc sharpness. With negative values small details get blurred a bit more, here’s my current preset using -10.00 ntsc sharpness:
Here’s with 10.00 sharpness:
With positive values the details like Sonic’s nose (the little white dots on it) stand out a bit more along with the details on the bricks on the wall and the fire lamp/fire itself.
Same thing with 3-phase games. Current preset with -10.00 ntsc sharpness:
Positive ntsc sharpness:
Notice the outer left and right long vertical black line on the “Round 1-1” board.
It’s very intricate and minor, some may not even notice the difference but it definitely helps if you’re going for a more sharpened look.
Thanks for some good test cases. Personally, i was always getting mad at the FF5 menu text with 3-phase. But that’s really more a chroma issue. Adaptive sharpen currently works on luma only, but i guess it’s a good idea to try it with chroma too some day.
Also a good solution would be tu run a multi-pass pattern checking shaders, very similar to the top dedithering solutions. I have narrowed it down to quite efficient lines of code, which still dissolve bars and checkerboards.
Dunno if it’s reasonable to look for complex solutions within these ntsc shaders, i guess there might be some possibilities left for improvements on the simple side, if something reasonable and decent shows up, i will sure update the shaders.
I also wanted to explain how lowering the number of taps also works as the chroma bleeding parameter would increase.
If number of taps is 16 out of total 32 (2-phase) then base bleeding value is already like 2.0, 4.0 becomes 8.0 etc. There is no need to crank 8 taps to 4.0, as it’s considered an extreme value and similar (like 16.0). This might only work with merge or s-video ‘presets’, just to mention it as the artifacted side effects are already mitigated as far it can go.