Jump to content
IGNORED

Graph2fnt


emkay

Recommended Posts

Are you sure about the DLIs?

With 5 colors you are usually in char mode and these complex images require to switch the charset from time to time, performed by DLI...

 

Busted! You're right. "No additional DLIs to change colors." It looks the DLIs range from as often as every 16 lines on the widescreen images to much fewer (dynamically) for images with lots of repetition or unused space.

Link to comment
Share on other sites

Busted! You're right. "No additional DLIs to change colors." It looks the DLIs range from as often as every 16 lines on the widescreen images to much fewer (dynamically) for images with lots of repetition or unused space.

But 5 colors could be done without DLIs, for screens, during SIO. Just put the PMg behind the playfield...

  • Like 2
Link to comment
Share on other sites

 

You tempted me also to try a quick 50% resize and I left all the normal problems uncorrected. The image still looked quite good at 1/4 screen. In this one, the face is a bit wider than 32 pixels, though.

 

DARK MATTER FIVE AT QUARTER-SCREEN:

attachicon.gifDMFive-mini.png

 

 

Thanks, that looks almost same as to what I got after resize. For me, the skin face was exactly 32 pixels.

 

I found a good rasta pic showing what I meant:

post-19882-0-20393600-1519240407.png

 

If you look at each side of the face, there's about 8 colors making the transition very smooth and natural. I see two major limitations without DLIs:

- we couldn't have those same colors at different xpos, so all those fine jittery details will be gone

- we can't have that transition on both sides of face - so the lighting would have to be just like with the five (from right side)

 

Still, a much improved visuals can be achieved even without DLIs if we somehow place&mask the PMGs carefully at that light boundary. I know I wouldn't like to mess with that via code, though...

 

Has this been attempted in past with faces ?

  • Like 1
Link to comment
Share on other sites

Here is another 5 color image of the team that everyone seems to love and loathe in equal parts. These are NTSC colors.

 

I feel a little guilty using Graph2Font just for converting, editing, and displaying an otherwise ordinary image that doesn't use any of Graph2Font's special tricks. I also kept the dithering to a minimum because it was introducing too much noise into the image. Three shades of gray, two shades of a matching skin tone. I probably should have used something like $26 and $2A for the skin, but I used the $3x range instead.

 

post-18231-0-01126400-1519636339.png

thetrams.xex

 

Here is the original image (after being scaled down and with the background removed):

post-18231-0-57981100-1519637693.png

 

Here is the target for importing into Graph2Font (which needed some manual corrections to remove flesh tones from clothing):

post-18231-0-42369900-1519636765.png

 

Once I took another look at the target image, I started to wonder if going for more a brown color looked better. What do you think? Does this recolored version look better or worse than the original? (NTSC only, please. If you view the binary, they'll look wrong in PAL, but this particular version will look far worse than the other one.)

 

Recolored image and binary:

post-18231-0-86943800-1519637480.png

thetrams-recolored.xex

 

Please enjoy (or seethe in anger if that's your thing),

jmccorm

 

EDIT: I had done this same image earler in RastaConverter. By comparison, the RastaConverter image had a lot better colors and it felt like it captured a lot more detail, but the version on this page is much more cleaner and has far less visual errors and noise. It just goes to illustrate some of the trade-offs which you might use, depending on the goal that you're trying to achieve.

Edited by jmccorm
  • Like 2
Link to comment
Share on other sites

A flawed (but informative) NTSC interlaced blending test. For the ladies?

 

Altirra Screenshot, 160x240 mode, ~13 pseudocolors (10 substantially used) with NTSC Color Blending:

post-18231-0-69928400-1519718067.png

 

Description:

I noticed that Photoshop and GIMP were producing some similar but different results in pre-processing images that were imported into G2F. So I took an image and pre-processed it with both tools, and then converted them into G2F images. Next, I exported the G2F files as an "animation", and I adjusted the code to set the delay rate of 0 frames between images, and I had to prevent the code from constantly decompressing the next image on every frame (which ate too much time). I used Altirra's Performance Analyzer to confirm that it was consistently interlacing the images between each pair of frames.

 

The interlaced G2F Images, slowed down, demonstrating the technique (and emphasizes the flaws):

post-18231-0-93753300-1519719480.gif

 

Observations:

ALTIRRA: Blends very well with System -> Video -> Frame Blending enabled. Looks great! Without that feature enabled, the larger square (left of "good" and "bad") either flickers or shifts between the two triangle halves. The photorealistic of the picture may flicker a bit, too..

 

CRT Television (Sharp): The text (top) jumps back-and-forth, there is some flicker on the photorealistic part of the image due to a few areas of significant differences in the preprocessed images, such as the eyebrow on the left. The largest square is very flickering, but a solid color, with one major exception. There is a bright line inside the square that goes from the lower left to the upper right, diagonally slicing across the middle.

 

Question:

I think I'll be creating a few more test patterns to figure out exactly where the boundaries are in interlacing two G2F images. (This interlacing is actually a known mode, but I forget which one.) I'm interested to know how well the various parts work on the screen(s) of people such as yourself?

 

Atari 8-bit Executable:

NTSC-Blend-Test-Flawed.xex

Edited by jmccorm
  • Like 1
Link to comment
Share on other sites

Super IRG mode. :) This is the same technique I have been using in my experiments, swapping two g2f renders every frame. Many of them involved swapping Antic 4 with a gtia mode, usually gr. 9 or 10.

 

You can process this with checkerboard dithering to reduce the flicker even further. Do you have the original image? I might give it a try.

  • Like 2
Link to comment
Share on other sites

Super IRG mode. icon_smile.gif This is the same technique I have been using in my experiments, swapping two g2f renders every frame. Many of them involved swapping Antic 4 with a gtia mode, usually gr. 9 or 10. You can process this with checkerboard dithering to reduce the flicker even further. Do you have the original image? I might give it a try.

 

Emphasis on the exact meaning of checkerboard dithering? Got it.

 

I remember you mentioning something along these lines in the RastaConverter thread, but I wasn't aware of your experiments. So I went back and took a better look in the programming forum. Wow! You've got some hefty expertise in this, and years of experience. Clearly, I am a novice. Actually, that's a good thing because it means that there is someone so much smarter that I can quickly absorb new tricks off of! icon_smile.gif

 

Of my many initial questions, I'll start with the following which are relevant to the previous image --

 

If I want to perform a monochromatic Super IRG (which is what the above image is a sub-optimal version of), how do you suggest that I...

  1. Use the 5th color? Does one typically leave it unused, or do they save a pair of them for a special purpose? (Many of the errors seen in this test message were from 3rd/4th color conflicts.)
  2. Assuming I go with static assignments for the colors (no DLI changes), what colors should I allocate to each pair of 4 or 5 registers? I figured there might be some clever mathematical formula to obtain the optimum intensity (or color) response? (That's potentially a tie-in to question #5.)
  3. If appropriate, should I shy away from the highest intensities to mitigate some of the potential for flickering?
  4. Is there a good rule for how far the difference between two frames can be for a pixel without the flicker becoming bothersome? Do you do anything to enforce these between the two images?
  5. Assuming my pre-processing has resulted in a 160x240 grayscale image, what is my next step if I'm going to split each individual intensity into the mathematical combinations of two sets of overlapping of 4 (or 5) colors? Or is that not how you're assigning the brightness for the two frames? Are you side-stepping the problem of optimizing the apparent intensity pre-processing to 320x240 and then horizontally interlacing a pair of pixels?
  6. If the original image is of significantly higher resolution, are you doing anything special to do some supersampling? Or do you just rely on the standard resampling filters in the resize function for that task?
  7. Still on the subject of supersampling: since I am initially doing a monochromatic image, I was considering for each pixel to do a 3-point sample of the original image for each 1 pixel in the final output. The three points would be positioned equidistant from the pixel center and arranged like a triangle (both an X and Y offset). Not quite a triangular subpixel arrangement because they'd be in the same physical space, and I'm thinking that I could get away with it because I'm not doing RBG. (But without a set of rules like from question #4 above, the chance of flicker might increase.) The subpixels would be displayed in frames 1-2-3-1-2-3 insteadl of 1-2-1-2. Interesting idea? Horrible idea?

EDIT: Is this temporal supersampling?

 

EDIT2: If I've got three frames, I could do something different. I could go 1-2-3-2-1 and compute frame 2 as the average of frame #1 and #3. This would be for the purpose of reducing/eliminating flickering of pixels which large differences between frames 1 and 3. But that could muddy the middle values, which would be unfortunate. It might be possible to offset that with a correction in the opposite direction as a 4th frame, but at that point, it is probably becoming needlessly complex. The effort is probably better spent on selecting appropriate intensities at the same pixels for each frame.

 

A less related question...

 

If I wanted a monochromatic 320x160 graphics mode, would you suggest Antic Mode 4 and Graphics 0? If so, what would be your choices for the Graphics 0 background intensity (plus appropriate color), and then the foreground intensity? 0% and 100%? 0% and 50%? 33% and 66%?

 

I bombarded you with questions. Sorry about that. It is nice to have someone with a wealth of knowledge and experience to lean on. I'll be sure to go back and pour over your messages in the programming forum. Regarding the image you asked about, I've got a high-resolution PNG, and then a PSD with the layers and with the background removed, and then I've created a PNG with the background removed just in case you can't easily work with PSD files. The catch is that they're all monochrome, so that may not have matched your expectations, but I'm hoping that doesn't throw too much of a wrench into your plans. I'll PM them here shortly.

Edited by jmccorm
Link to comment
Share on other sites

Well, before I get started ...

 

There are two ways that I have used Super IRG in my experiments ... that is, an Antic 4 screen with font changes. Typically you don't alter the color registers at all, just the font data and screenmap. This allows for 14 (or 15 - I will explain later) colors usable, in theory. However, when doing greyscale, you don't really have all of these. It's more like 9 level shading.

 

If I want to perform a monochromatic Super IRG (which is what the above image is a sub-optimal version of), how do you suggest that I...

  1. Use the 5th color? Does one typically leave it unused, or do they save a pair of them for a special purpose? (Many of the errors seen in this test message were from 3rd/4th color conflicts.)

 

 

You can ... when blending two charactermaps though, you do have to watch out for PF2/PF3 conflicts alot. You can't have PF2 and PF3 in the same character cell in the same frame. If you make two screenmaps, and flip between them, PF2 and PF3 blending becomes possible (getting you a 15th color) but the tradeoff is those areas of the picture can't be checkerboard dithered. One possible method that I haven't tried yet, is making two G2F renders, one with 5 colors, and one with only 4, and blending the two to make a sort of 9-level greyscale picture.

 

There was a bitmap mode called Interpainter that did this sort of thing. You blended two Graphics 15 bitmaps to get a 7-greyscale monochrome picture. Atari Interlace Studio supports this mode.

 

Assuming I go with static assignments for the colors (no DLI changes), what colors should I allocate to each pair of 4 or 5 registers? I figured there might be some clever mathematical formula to obtain the optimum intensity (or color) response? (That's potentially a tie-in to question #5.)

 

 

I always allowed G2F to set the color palette ... In general, having your darkest shade as PF2, and your brightest shade as PF3 (or vice versa) tends to avoid a lot of conflicts. The colors should remain steady in both frames for best results. However, if you do change the colors (a mode I named Super IRG+ or Super IRG 2) you can increase your color resolution, but it means you must do full frame changes. So if you were to change your colors, you could try something like this:

 

Frame one: PF0: 6 PF1: 10 PF2: 2 PF3: 14

Frame two: PF0: 8 PF1: 12 PF2: 4 PF3: 14

 

PF3 will be the same in both frames, because the Atari only has 8 luminances in Antic mode 4. If you are using the SOPHIA video card, you can use 16 greyscale in Antic 4.

 

For pictures that don't change the colors each frame, maybe a setting like this:

 

PF0: 6 PF1: 10 PF2: 2 PF3: 14

 

  1. If appropriate, should I shy away from the highest intensities to mitigate some of the potential for flickering?
  2. Is there a good rule for how far the difference between two frames can be for a pixel without the flicker becoming bothersome? Do you do anything to enforce these between the two images?

 

 

Well the key is, you don't want the pixels being blended to differ than more than 2 steps luma. High luminances are not bad in and of themselves, but in greyscale pictures it limits the number of actual colors you get. In theory there are 14/15 color combinations in Super IRG (23 to 25 in Super IRG 2 with color changes) but combinations which blend the highest and lowest luminances should not be used ... so in actuality, you would end up with about a 9 level greyscale picture in Super IRG, or maybe about 17 level in Super IRG 2.

 

So for example, in Super IRG (no color changes), these would be your 9 colors:

 

0, 0/2, 2, 2/6, 6, 6/10, 10, 10/14, 14 ... keeping in mind that you can't have lum 14 and lum 2 in the same character cell as these are PF3 and PF2 respectively.

 

I haven't done a lot of monochrome pictures in these modes, as I have never been happy with the results. One mode that I have gotten good monochrome results out of is one I named MIN (Monochrome Interlace) ... you blend an Antic 4 screenmap with a Antic 2/GTIA 9 (16 grey) screenmap. This mode gets you about 20-21 greyscale but with a lot less flicker, and the 160 pixel resolution is still apparent. I generally prepare my pictures in GIMP and resize them to a 320 wide resolution. Then you can either rescale them to 5 colors, or just use G2F to rescale them. My MIN pictures are a straight blend of the G2F's created in Antic 4 and GTIA 9 respectively. I'm going to try a MIN render of your picture to see how it comes out.

 

If I wanted a monochromatic 320x160 graphics mode, would you suggest Antic Mode 4 and Graphics 0? If so, what would be your choices for the Graphics 0 background intensity (plus appropriate color), and then the foreground intensity? 0% and 100%? 0% and 50%? 33% and 66%?

 

 

 

Antic 4 and Graphics 0 (a mode I have named DIN, Display Interlace) is a good combination. There are some requirements to make this work though:

 

1. You have to alter PF2 each frame. In the Antic 2 frame, it should be toggled to the same value as BAK to avoid background flicker.

 

2. You may need to alter PF1 each frame. PF1 works best in this mode at about luma 10, but you may want to play with it. You will then either need to arrange the palette in Antic 4 so that the brightness 10 color you use is at PF1, or you will need to alternate the color each frame.

 

3. You will need to either flip two screenmaps (with the screenmap in Antic 2 having no inverse characters), or alternate CHACT to kill inverse video when the screen is showing Antic 2. Otherwise the characters where PF3 is present in Antic 4 will show garbage in Antic 2.

 

What I usually do is process the Graphics 0 render in GIMP and apply Floyd Steinberg dithering, this gives the illusion of greyscale. Keeping in mind, on a CRT you will get color artifacting which can muddy the picture. There are some DIN pictures in my thread which show what can be done.

 

I've also done the same by blending Graphics 10 and Antic 2 (a mode I called PC0 or Palette Color 0) ... it does ok, depending on the picture, but the 4x1 pixel blending tends to look a bit muddier than DIN mode.

 

Hopefully this proves to be of help.

  • Like 1
Link to comment
Share on other sites

OK here's an attempt:

 

post-23798-0-26407400-1520038979.png idude-min.obx idude-min.asm

 

 

I did this in MIN mode, quick and dirty using a graphics 9 and a graphics 12 source g2f. It's not optimal, but it gives you an idea.

 

To optimize this best with minimal flicker (better than I did above), you would want to convert your original image to 21 level greyscale. Then use these settings for Antic 4:

 

BG - 0

PF0 - $06

PF1 - $0A

PF2 - $0E

PF3 - $02

 

Then arrange your palette as so to get your 26 greys:

 

Antic 4 Graphics 9 Combined greyscale

BG - 0 0 1 2 0 1 2

PF3 - 2 0 1 2 3 4 2 3 4 5 6

PF0 - 6 4 5 6 7 8 7 8 9 10 11

PF1 - 10 8 9 10 11 12 12 13 14 15 16

PF2 - 14 12 13 14 15 17 18 19 20

 

BG/2 and 2/0 are the same color and can be checkerboard dithered.

 

Then read the pixel pairs of your 21-grey PNG and convert them to the paired Antic 4/Graphics 9 colors using the table above:

 

You might need to rearrange your Antic 4 palette as to best avoid PF2/PF3 conflicts in the picture but in theory this will provide an optimal picture.

Edited by Synthpopalooza
  • Like 1
Link to comment
Share on other sites

Also, your Tramiels ... in PCIN mode:

 

post-23798-0-71895000-1520042295.png trams-pcin.obx trams-pcin.asm

 

I used a different source G2F than your render ... and I also converted the picture to Graphics 9 colors using GIMP

 

post-23798-0-30287300-1520042314_thumb.png - Graphics 10 g2f

 

post-23798-0-97487300-1520042334_thumb.png - Graphics 12 g2f

 

There are optimizations that could still be done, for example, where the PF0-PF3 and BAK colors intersect in the picture, checkerboard dithering can be done to improve the flicker.

 

 

 

 

 

  • Like 1
Link to comment
Share on other sites

I would have replied sooner, but you've got so many awesome things for me to latch onto that I really wanted to do a reply justice. The most immediate thnk that popped out was how to easily create a checkerboard dither inside of Gimp or Photoshop if you're doing something as similar a frame-blending a fixed-palette set of Antic 4 images. It doesn't seem like you can just invert the pixels and still have it work out, and there isn't an easy was that I know of to tell Photoshop to do an inverse checkerboard, so I'm trying to think of if there is a good strategy to do this by hand (and not actually program out some kind of conversion tool).

 

Overall, you've given me a ton of stuff to work towards. My own personal preference, though, is probably going to be to stay away from the 4 color clock modes, because I think that the 160x modes are bad enough as it is. In fact, I'd probably want to do the opposite direction and blend with an Antic 2 if it bumps up the detail. Being limited by resolution has been far more frustrating to me than getting ahold of a large number of colors. Case-in-point is my current work-in-progress, a picture of an Atari 800. I just don't have enough resolution to pull off the Atari 800 nameplate which is printed at an angle, in a small area, and with 2x1 pixels. (Also those 2x1 pixels just seem to make dealing with jaggies even harder.)

 

post-18231-0-37873800-1520064982.png

 

IMAGE FILES:

atari-800-g2f-source.zip

atari-800.xex

 

BTW, I know this pretty much rules out interlacing, but did you ever do some public experiments with the 480i soft modes?

 

 

  • Like 2
Link to comment
Share on other sites

Title: Atari Age Desktop Logo

Desktop background created by: RevEng in October, 2013

Converted to 8-bit binary by: Jmccorm in March, 2018

Alternate title: "It never hurts to suck up to the boss."

 

The image is a combination of Graphics 0 (320x192 with 2 colors) and Antic 4 (160x192 with 5 colors). The top 2/3rds of the screen uses Graphics 0 for the extra resolution but compensates for the lack of color choices by using Player/Missile Graphics (at a lower resolution) to obtain a white logo. Antic Mode 4 filled out the bottom of the display (where resolution was less important) and it was merged together on a large overscan playfield. Color palette and aspect ratio were optimized for NTSC, no dithering or interlacing (this one didn't need it). I considered applying a standard rainbow color cycle for the AtariAge nameplate, but I didn't think it would elevate the image. There are plenty of unused resources available for other enhancements, however.

 

Again, the original AtariAge desktop image was made by RevEng. All I did was harvest it for the experience in working with Atari 8-bit display specifications. I'm creating some small pieces here and there to exercise some of the skill necessary for my own project. Just to make sure I'm stretching the skillset in a well-rounded way, it probably wouldn't be a bad idea for me to entertain a few requests.

 

IMAGES AND ATTACHMENTS:

post-18231-0-96282500-1520154193.png

AtariAge.zip (G2F Source File)

AtariAge.xex

Edited by jmccorm
  • Like 7
Link to comment
Share on other sites

Title: Atari Age Desktop Logo

Desktop background created by: RevEng in October, 2013

Converted to 8-bit binary by: Jmccorm in March, 2018

Alternate title: "It never hurts to suck up to the boss."

 

 

Oh no! This reminds me of the (red+white) japanese war flag.

 

So now we have the AA war flag on our desktops ?!?

Edited by CharlieChaplin
  • Like 1
Link to comment
Share on other sites

Have never played around with the 480i mode ... it requires a lot of overhead but it does bear some experimenting perhaps.

 

I did an example for you to look at ... DIN mode (Antic 2 + Antic 4) plus source code.

 

Some notes: For the High res Graphics 0 render I took the image into GIMP, and posterized it using a 1-bit binary palette, with Floyd Steinberg dithering. Save as PNG and import it into G2F to get font and SCR files.

 

In the code, I set it to where PF1 and PF2 are altered every scanline. Where the screen is Antic 2, PF2 should be the BAK color and PF1 should be $0c (brightness 12 decimal) though it may be necessary to play with these settings to find the optimal picture with minimal flicker. Where the screen is Antic 4, these two settings will be the normal colors for the Antic 4 render.

 

post-23798-0-58008700-1520233658.png spaniel-din.obx spaniel-min.asm

 

ZIP file with source g2f's:

spaniel source g2f.zip

  • Like 2
Link to comment
Share on other sites

I would have replied sooner, but you've got so many awesome things for me to latch onto that I really wanted to do a reply justice. The most immediate thnk that popped out was how to easily create a checkerboard dither inside of Gimp or Photoshop if you're doing something as similar a frame-blending a fixed-palette set of Antic 4 images. It doesn't seem like you can just invert the pixels and still have it work out, and there isn't an easy was that I know of to tell Photoshop to do an inverse checkerboard, so I'm trying to think of if there is a good strategy to do this by hand (and not actually program out some kind of conversion tool).

 

Overall, you've given me a ton of stuff to work towards. My own personal preference, though, is probably going to be to stay away from the 4 color clock modes, because I think that the 160x modes are bad enough as it is. In fact, I'd probably want to do the opposite direction and blend with an Antic 2 if it bumps up the detail. Being limited by resolution has been far more frustrating to me than getting ahold of a large number of colors. Case-in-point is my current work-in-progress, a picture of an Atari 800. I just don't have enough resolution to pull off the Atari 800 nameplate which is printed at an angle, in a small area, and with 2x1 pixels. (Also those 2x1 pixels just seem to make dealing with jaggies even harder.)

 

attachicon.gifAtari-800.png

 

IMAGE FILES:

attachicon.gifatari-800-g2f-source.zip

attachicon.gifatari-800.xex

 

BTW, I know this pretty much rules out interlacing, but did you ever do some public experiments with the 480i soft modes?

 

 

The resolution is what it is. If I were attempting this picture, I would "zoom" in on the 800 computer so it completely fills the screen, possibly giving you enough pixels to improve the nameplate. Personally, I think it looks fantastic as is.

  • Like 1
Link to comment
Share on other sites

  • 2 weeks later...

I'm redoing a RastaConverter image of a classic Battlestar Galactica Colonial Viper. I thought that it was a really cool image. My run is currently is at over 1 billion iterations and I'm letting it continue overnight. In the meantime, I thought I'd try a hand at doing it with G2F. I think it turned out pretty good, and in this case, it will probably be better than the Rasta version. Regardless, I still intend to put my Rasta version out there tomorrow.

 

The reason that the G2F version is better, I think, is that this is a picture of a mechanical object where details matter and colors are less important, so that puts Rasta at a disadvantage. (Hey, totally not knocking on RastaConverter. I love it! But different tools are going to have advantages and disadvantages in certain areas.) One of the disadvantages in using G2F is that I had to do a lot of prep work to feed it a good image, and then I had to draw back in the red stripe by hand (as a 6th and 7th color to match the brightness of the surface).

 

This is a wide-overscan image designed for NTSC colors and aspect ratio. NTSC users on CRTs probably won't see the dynamic video garbage on the far right. I wanted to mask that with a PMG object, but since my picture used GTIA Priority mode 4, I didn't have the priority I needed (mode 1) to mask the garbage, and there wasn't any CPU left over for DLI tricks.

 

Anyhow, enough talk. Here it is.

 

PICTURE:

post-18231-0-60706700-1521444404.png

 

BINARY:

Classic-Viper-Jmccorm-G2F.xex

 

ARCHIVE WITH G2F SOURCE, IMAGE, AND BINARY:

Classic-Viper-Jmccorm-G2F.zip

 

It looks pretty good for a quick effort image on an Atari 8-bit?

If they only had better tools and access to more information back in the day.

 

Enjoy. Rasta version should arrive tomorrow.

  • Like 8
Link to comment
Share on other sites

Minor modification:

If the dynamic garbage on the right bugs you (usually when viewed on a widescreen display or PC), I converted the image to work without the the wide playfield. I shifted the entire image to the left, and then I used two colors of PMGs to re-create the wing in the left-hand border area.

 

What I learned:

If you are using two different colored PMGs to paint the same area, and you've got one PMG at normal width and one PMG at double-width, you should try to arrange it so that the normal width PMG has priority over the double-width PMG, so that its finer resolution could be used to mask to be used against the course double-width PMG. The exercise was a good learning experience.

 

BINARY:

Classic-Viper-Jmccorm-G2F-Half-Overscan.xex

  • Like 1
Link to comment
Share on other sites

I like to make some comment on the Feb 26 post - with the Tramiels group artwork done...

 

I presume with the Graph2fnt program - that with the 5 colour mode - It is possible to do some hand touch ups on the image. If there was enough pixels present - you could touch up the eye area(s). Of course, there's the lack of pixels to provide any decent detail there - but you could possibly? hint the presence of detail.

 

If you wanted to do the best possible - you simply need to re-arrange the grouping so that you would have more pixels to work with. ie. Place the figures closer together such that they could look natural together (instead of being an obvious composite of separate images).

Maybe you considered this already - and simply did not want to go through the process (and time) to do this...

 

What you've done is of a high standard already - so disregard my commenting - if you wish to.

 

Harvey

Link to comment
Share on other sites

I like to make some comment on the Feb 26 post - with the Tramiels group artwork done...

[ ... ]

What you've done is of a high standard already - so disregard my commenting - if you wish to.

 

Wow! I guess I have two different responses to that?

 

First, I was first impressed by your artwork when I experienced it back in the day (from a series of demo disks which hailed from New Zealand). Any of your observations are most welcome! Second would be that, in general, I do take most comments quite well as long as they aren't coming from a position of malice, which yours most certainly are not.

 

After I originally posted the image, later that day, I worked on it just a tiny bit more. I did some stray pixel cleanup, made Jack's head a little less round, and whitened some of the smiles. Here is what I ended up with...

 

post-18231-0-83371400-1521516419.png

TheTrams-retouched.xex

 

If you wanted to do the best possible - you simply need to re-arrange the grouping so that you would have more pixels to work with. ie. Place the figures closer together such that they could look natural together (instead of being an obvious composite of separate images).

 

Maybe you considered this already - and simply did not want to go through the process (and time) to do this...

 

Actually, as far as I could tell, the arrangement wasn't manufactured. The source appears to be an original photo from the period? Here it is:

 

NOTE: IMAGE EDITED AFTER POSTING

post-18231-0-58321400-1521562413.png

 

March 20th, 2018: After researching another user's request for a higher resolution copy, I found that Getty Images may own some rights to this image. Their user agreement allows personal use (generally defined as non-commercial behavior) of images from their website, which I am doing in my adaptions. Yet I don't think they'll want to see even medium resolution copies of their raw images embedded in a forum. Out of an abundance of caution, I have cropped and blurred the above image. If anyone is interested in obtaining a high quality copy, they sell a 3072x2048 image for $575 on their website. They have other (and more restrictive) commercial and non-profit licensing options available at somewhat reduced costs. -jmccorm

 

In general, I find that I am able to make almost halfway decent artistic observations, yet I consider my actual artistic skills to be very novice in nature. My experience is low but improving. I'm glad that these images are turning out well enough to earn some praise. I think that both the limitations of the medium as well as the great tools really help out.

 

With some effort, I could probably pull Garry, I think it is (upper left), and reposition him in the lower right. But then I'm recomposing a real photograph (and with my uncertain artistic skills and limited colors, I'm not sure how I'd handle his black hair against the silhouettes of his family members). Then I could pull Sam (lower left) in a bit more towards the center and further up so I could get a tighter crop. But it is a lot of work. You're totally right, though, that too few pixels are being spent on actual faces, and more on attire. I'll take that as a lesson, going forward.

 

I presume with the Graph2fnt program - that with the 5 colour mode - It is possible to do some hand touch ups on the image. If there was enough pixels present - you could touch up the eye area(s). Of course, there's the lack of pixels to provide any decent detail there - but you could possibly? hint the presence of detail.

 

Graph2fnt allows for a great deal of image editing (within a fairly good subset of what the Atari 8-bit makes possible). It goes beyond just being a pixel editor. PMG objects. DLI color and PMG changes on one or more lines. Mid-line register changes. Mixed graphics modes. Variable playfield sizes. I'm sure that I'm missing some things. Once you get the hang of it, it offers plenty of opportunities to improve a basic image.

 

I'll run another conversion and see if I can better capture some of the whites of the eyes and see if I can use that to inform a few alterations. You're right, their eyes are both small and dark, and I left their eyeballs as a skin tone. It really is asking for some improvement, and aside from the three shades of gray, I've got plenty of PMG resources that I could spend on it. But we're reaching the limit of pixels in the source image. Aside from that, I should probably add some gray back into Jack's hair, rather than presenting him as bald on the sides.

 

I'll make these touch-ups one of my next challenges, thanks!

(And it is always good to hear your point of view.)

 

EDIT: Come to think of it, I could reprocess the entire image into a 5 level grayscale, and with careful work (and probably putting Leonard in the upper right to be just below Jack), I could use some of the PMGs in priority mode 0 to get color-mixing to give us back some shaded flesh tones, where needed? All of this is actually why I like reading comments on these images. I' want to grow my digital artistic skillset!

 

-jmccorm

Edited by jmccorm
  • Like 2
Link to comment
Share on other sites

I can't do the repositioning myself - like some virtual photographer so as to move them into a tight frame so to speak - because I use only Paint to a very limited degree - but maybe someone else can do that? Who knows it's a simple matter for them to do so.

But alas you also need some artistic talent to create some detail where there is none or very little.. who is familiar with using Graph2fnt.

Fitting 4 heads into a horizontal frame is not going to be an easy task - will it look nice? But it will provide the extra pixels in which the all important eyes can be rendered better maybe?

This kind of approach might be useful for some other future picture ...

 

It's always good to see people getting involved and having a go with stuff, graphics and whatever else is their interest, etc.

That you're not alone in wanting to do something nice on/for this 80s' hardware - that was ahead of it's time - and deserved better recognition than what it got..

 

Harvey

  • Like 2
Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...