Wow, Britney. Bravo!Now let's touch your soul.
DGDenoise
DGDenoise
DGDenoise
Skimmed all the pages & i'm trying to figure out if this is good for getting rid of fairly heavy grain or not?
Messed around with Strength a bit but it kills detail if you go hard & idk what the other parameters really do in the technical sense, i did read the manual but this isn't my battlefield.
Can someone maybe give me a...
DGDenoise (strength=xxxx, cstrength=xxxx, blend=xxxx, cblend=xxxx, searchw=x)
...set of parameters to jump off. I hate grain.
Working with 4K YUV 4:2:0 10bit
Messed around with Strength a bit but it kills detail if you go hard & idk what the other parameters really do in the technical sense, i did read the manual but this isn't my battlefield.
Can someone maybe give me a...
DGDenoise (strength=xxxx, cstrength=xxxx, blend=xxxx, cblend=xxxx, searchw=x)
...set of parameters to jump off. I hate grain.
Working with 4K YUV 4:2:0 10bit
DGDenoise
Hi Whackbag. Can you post a link to an unprocessed source stream sample? It's really only guesswork without that. Thank you.
DGDenoise
I guess it wasn't that important to him after all.
Curly Howard
Director of EAC3TO Mod Development
Director of EAC3TO Mod Development
DGDenoise
Just curious.
Would process DGDenoise "inside" DGSource speed up the processing?
In my naive mind it seems that the frame is copied to system memory then back to gpu one to be processed.
Is it like that? The question applies, of course, to any other CUDA filter.
Would process DGDenoise "inside" DGSource speed up the processing?
In my naive mind it seems that the frame is copied to system memory then back to gpu one to be processed.
Is it like that? The question applies, of course, to any other CUDA filter.
DGDenoise
Hey big t.
Curiosity killed the cat. Not sure about frogs.
Yes, of course, and the transfer overhead is massive for UHD. This is the whole point of the stillborne CUDASynth project. Nobody showed even an iota of interest in that, even though I demonstrated performance gains of 300%+ for typical scripts. Efforts by others in this regard were also abandoned. I don't get it. Maybe as 4K and 8K take over things can change.
Sure, I can put DGDenoise() "inside" DGSource(). I can put lots of things inside. However, I don't see any significant usage of these filters such as to justify the not inconsiderable effort. But maybe.
Anyway, I got an itch today to allow the DG source filters to accept a media file name in place of a DGI file name. Of course a DGI file name would still be accepted. Then, if a media file name is given and the corresponding DGI file does not exist, it would be automatically indexed. If the DGI file already exists it would simply be used. I believe some competing source filters work this way and several people have asked for this behavior. My only question is how to tell the user that index creation is taking place so they don't wonder what the heck is going on with the long delay. Do you know what our competitors are doing in that regard?
Sorry for the OT, but I'm not going to look at my feet.
Curiosity killed the cat. Not sure about frogs.
Yes, of course, and the transfer overhead is massive for UHD. This is the whole point of the stillborne CUDASynth project. Nobody showed even an iota of interest in that, even though I demonstrated performance gains of 300%+ for typical scripts. Efforts by others in this regard were also abandoned. I don't get it. Maybe as 4K and 8K take over things can change.
Sure, I can put DGDenoise() "inside" DGSource(). I can put lots of things inside. However, I don't see any significant usage of these filters such as to justify the not inconsiderable effort. But maybe.
Anyway, I got an itch today to allow the DG source filters to accept a media file name in place of a DGI file name. Of course a DGI file name would still be accepted. Then, if a media file name is given and the corresponding DGI file does not exist, it would be automatically indexed. If the DGI file already exists it would simply be used. I believe some competing source filters work this way and several people have asked for this behavior. My only question is how to tell the user that index creation is taking place so they don't wonder what the heck is going on with the long delay. Do you know what our competitors are doing in that regard?
Sorry for the OT, but I'm not going to look at my feet.
DGDenoise
Hehe, I made a test version of the auto-indexing. Let's continue discussion here:
https://www.rationalqm.us/board/viewtop ... 807#p18807
https://www.rationalqm.us/board/viewtop ... 807#p18807
DGDenoise
We can live in rainforest, there is little if any that can kill us or stand our poisonous skin.
As you know, the only person who is maintaining a lot of AVS+ stuff is pinterf. He is getting old and time is becoming a valuable thing. Feel free (and be encouraged) to submit patches in github AVS+ to include proper CUDA support. The community would be so grateful!
Honestly, I always hoped for that. At least filters that can be applied without any other "external" intermediate. DGDenoise, sure, but DGDecimate and similar too. Who yelled DGCube?
Really interesting. Do you plan to extract other streams too while automatically indexing?
They tell plain nothing. The user waits until indexing has been finished.
I am not good at english idioms. If I understood properly, it's like the italian "to stay with hands in hands".
You know... I always dream of avs_resize on cuda or, at least, dgcube internal conversion.