JCC not working (empty images)

[jack_norton]

Today decided to try JCC, ran it and generated all the jpg images. However when I try to run the game, all converted images are empty (not displayed in the game). I don't know why this happen, any clues? the jcc.rpy file obviously is there, and the files as well.

[PyTom]

hm... can you attach a couple of the generated images? Specifically, the original png and the .mask.jpg and .base.jpg files generated for it? That will help me determine which step is going wrong.

[jack_norton]

Well for now I just used 256 color png. For manga style art is more than enough (you don't even notice the difference) and also the size is even smaller than using JCC

[Jake]

Well for now I just used 256 color png. For manga style art is more than enough (you don't even notice the difference)

Hmm. You quite possibly wouldn't notice the difference colour-wise, presuming you're talking about cel-shading, but... surely the difference is noticable in terms of transparency? Since 24bpp PNG and JCC can do varying levels of transparency, making the edges of the sprite smooth, while 8bpp (256-colour) PNG can only do on/off transparency and thus the edges will be abrupt and pixellated?

[jack_norton]

Well I am using 1024x768 as resolution, so is barely noticeable (at least for me). For now I'm going to leave it this way, can always change it later

[jack_norton]

Ok I've found the time to upload the images. They look fine to me (mask is ok), still wonder why I get nothing displayed ??

[jack_norton]

I even tried to load manually the files with:

image splash = im.AlphaMask(im.Image("gfx/logo.png.base.jpg", cache=False), im.Image("gfx/logo.png.mask.jpg", cache=False))

but still nothing is displayed. Nobody uses JCC ? I tried loading the images in my own C library and works, so must be a bug of renpy or SDL ?

[jack_norton]

I think I found the bug in renpy:

Function: im.AlphaMask (base, mask, **properties):

base and mask should be image manipulators. This function takes the red channel from mask, and applies it to the alpha channel of base to create a new image.

the mask image is a grey JPG, so can't take the red channel I think... ?

[Jake]

the mask image is a grey JPG, so can't take the red channel I think... ?

All pixels in a digital image are represented as red, green and blue values - even black and white. Typically, they'll range between 0 and 255, the maximum value you can store in a single byte.

Black is the absence of all colour, so it's RGB 0/0/0; White is all colour at the same time, so it's RGB 255/255/255. A 50% neutral grey is half-way between black and white, and as such it has an RGB value half-way between, of 128/128/128 (or 127/127/127, whichever you prefer).

So yes, a greyscale image still has a red channel, it's just that the red channel, the green channel and the blue channel will be the same. I expect that the reason the function uses the red channel in particular is just that it's convenient; as it goes, frequently the red value is stored in the lowest byte of a 4-byte integer for each pixel. For a greyscale mask image, you can read that sentence instead as "This function takes the luminance value from mask, and applies it to the alpha channel of base to create a new image."

[jack_norton]

Ok, thanks for the explanation. Still it doesn't work does anybody uses JCC successfully?

[PyTom]

It worked when I tested it, a while ago. I'll debug it when I get back home, mondayish.

[DaFool]

Ok, thanks for the explanation. Still it doesn't work does anybody uses JCC successfully?

The yaruge hentai I released early this year successfully used jcc, and it's open-source in case you want to take a look (and please try to ignore the general pervertedness and crappiness of the work otherwise)

Recently I'm just making graphics, so I haven't been fresh programming-wise for a while now (my mind goes blank on technical stuff nowadays).

What I do remember is make sure the png really needs the jcc functionality, i.e. it has transparency. Otherwise, in some instances, (and in virtually all cases where there is no transparency), a straight jpeg export is more efficient. In other instances, the deterioration of the jpeg mask is apparent.

What I did was compare pngs to jcc-induced jpegs and pick the ones with the best quality/size ratio. For images that have gradual gradients (such as, ahem, milky-white substances), you can't use jcc on the pngs without introducing pixelation around the edges, so you have to keep using pngs.

[jack_norton]

Thanks, I'm going to check it. It uses an archive or I can extract 2 test images to check?
(even if as I said before is really strange, in my own C code the 2 jpeg made with JCC works fine! so I really don't understand the bug )

[Gau_Veldt]

Thanks, I'm going to check it. It uses an archive or I can extract 2 test images to check?
(even if as I said before is really strange, in my own C code the 2 jpeg made with JCC works fine! so I really don't understand the bug )

Try making a project with a minimal script to display just the JCC'ed image. If it still doesn't work in that minimum project, zip that up and post it. Also make note of what version of Ren'Py you are using.

[jack_norton]

Well actually the images I posted use a minimal script
I'm using latest renpy. I use that image to replace the default window/buttons background, but even if I try to show a simple image on startup, doesn't work

Code: Select all

init:
    # Declare images below this line, using the image statement.
    image splash = im.AlphaMask(im.Image("gfx/logobase.jpg", cache=False), im.Image("gfx/logomask.jpg", cache=False))    #image splash = "gfx/logo.png"

label splashscreen:
    $ renpy.pause(0)
    scene black
    show splash at truecenter
    with dissolve 
    with Pause (1.0)

the logo images: