Images generated by RastaConverter

[Beeblebrox]

1 hour ago, Philsan said:

Very nice picture.

 

Would it be possible to avoid those lines?

At the time I tried to use Sheddy's tool to remove. In the end they are so small I didn't bother trying another attempt. Very happy with it as is.

 

Edited October 29, 2023 by Beeblebrox

[patjomki]

There is a tool in the slideshow generator that tries to fix these lines. Most of the times it helps 

[Beeblebrox]

1 hour ago, patjomki said:

There is a tool in the slideshow generator that tries to fix these lines. Most of the times it helps 

Yup, I mentioned that tool earlier.  I use it all the time and used it on all my conversions by default. 

[+Atari8man2004]

On 10/28/2023 at 3:45 AM, Beeblebrox said:

In part. I take the image from Irfanview after decreasing colour depth, resizing to 320x240, (or 200 if letterbox), and tweaking contrast and saturation, but you can use any similar image manipulation editor.

 

Then I paste it into paint.net. I use the latter's lasso select tool to cut out the shape around the image I want to mask, then use the fill in tool with the colour black to full in all those areas left which I don't want to mask in. Then I cut the entire image, (say 320x240), paste it back into Irfanview and save it as a jpg with the same name as the image with the word mask in the filename so I know which is which. I also usually reduced the number of colors of the mask down to 2, so it's just black and white, before I save. I just do that out of habit. 

 

RC then knows that the bit you cut out, in this case white, is the bit you want it to prioritise/focus on, and the other bits in black it doesn't. 

 

Experiment and you'll soon see.

 

Edit: @Atari8man2004 crude example as I've just done this on my phone with my finger in a very rudimentary app

Say you want the mask created to focus on the head in this image, this is how the mask would look:

 

 

 

Then in mask settings in RC gui I usually set it to 10. Experiment.

 

The more you priorise the masked area, the more any details outside of the mask will suffer as it were, but that depends on the source image and background. I tend to use wider masks that almost cover entire images. Sounds odd but it works. And for my images like the girl one, where the background in black, you can just cut the mask around her head, like in the above example. Just experiment. 

 

All my images in the past 12months use masks. Before that they didn't and it's clear some previous conversions would have been a lot better had I used one. I may revisit an early image and use a mask, to illustrate this. 

I have 2 conversion in the oven both have mask and I can see how masks handle areas that the converter had problrms with.  Should be ready in a few days 

[+Atari8man2004]

RT_2b   mask were used to enhance reflections.  My first time using a mask 

Atari8man_RT_2b.xex

[Beeblebrox]

16 minutes ago, Atari8man2004 said:

RT_2b   mask were used to enhance reflections.  My first time using a mask 

Atari8man_RT_2b.xex 20.79 kB · 1 download

Very nice!! I know it's asking a lot but I'd love to see it without any dithering.

 

Also, try either jarvis or Floyd dithering modes if you have to use dithering. Imho they look better. I tend to also crank up the dithering random setting. 

Edited October 31, 2023 by Beeblebrox

[+Atari8man2004]

1 hour ago, Beeblebrox said:

Very nice!! I know it's asking a lot but I'd love to see it without any dithering.

 

Also, try either jarvis or Floyd dithering modes if you have to use dithering. Imho they look better. I tend to also crank up the dithering random setting. 

Looks good without dither, back in the oven  uploaded the dest file below

Edited October 31, 2023 by Atari8man2004

[drpeter]

2 hours ago, Atari8man2004 said:

RT_2b   mask were used to enhance reflections.  My first time using a mask 

Atari8man_RT_2b.xex 20.79 kB · 4 downloads

Extraordinary.  I like the dithered version 😀

[+Stephen]

On 10/29/2023 at 9:17 AM, Beeblebrox said:

 

Pictures (and video) will be posted later this evening, showing the atrocious way my Pixel Pro 6a captures 50Hz CRT.  It's not possible with the standard camera app.  Sorry.

[Beeblebrox]

34 minutes ago, Stephen said:

Pictures (and video) will be posted later this evening, showing the atrocious way my Pixel Pro 6a captures 50Hz CRT.  It's not possible with the standard camera app.  Sorry.

No worries, thanks for trying. 

[+Stephen]

21 minutes ago, Beeblebrox said:

No worries, thanks for trying. 

It's OK - I will post so that maybe I can get some help or suggestions.  I have some time off work coming up, I can try my wife's proper digital camera on a tripod and maybe get better results.  Failing this, I can try getting images from an NTSC machine, but that of course will throw the palette off on many.

[+Atari8man2004]

Sifi_Lady - Mask greatly improved image.

Atari8man_Sifi_lady.xex

[+Atari8man2004]

Artificial intelligence 2 - Computer generated art.

 

Atari8man_artificial intelligence 2.xex

Edited November 2, 2023 by Atari8man2004

[+Atari8man2004]

On 10/28/2023 at 11:27 AM, Beeblebrox said:

I think I learnt this from Amarok. 

This also enables you to do the double dithering I've recently alluded to. Double dithering sounds like something I increasingly find myself doing as I age heh heh!!

 

Seriously though doubling up on dithering is pretty effective.

 

So in the first pass you use dithering, (my preference is either flloyd or jarvis), then get the destination image to then make that the source, and then use dithering again. Had some cracking results that way.

 

Rc is all about experimenting, tweaking, experimenting some more, tweaking, etc etc. Defo the more prep you do and the more tweaking of the gamma, contrast and brightness you do in rc settings, previewing each time, the better. I can ramp up the latter settings, say, on the first pass to create the destination file I am gonna then use as the source, then almost find it needs the settings to come back down. The preview function more or less allows you to fine tune it, although as all know you don't always end up with exactly the right result. 

 

Between all this and masks, and of course being savvy with the actual image you choose to convert in the first place, you can get some amazing results. I was so pleased with girl portrait image (again below). This was done that way and the double dithering really enhanced the skin tones, etc imho.

 

Again I'd love to see this on real hardware on a CRT.

Send me the link and I'll load it on a crt.

[Beeblebrox]

1 hour ago, Atari8man2004 said:

Send me the link and I'll load it on a crt.

Hiya. Thanks.

Here it is again (I think you missed it the other day ;)) 

https://forums.atariage.com/topic/200118-images-generated-by-rastaconverter/?do=findComment&comment=5299136

 

[+Atari8man2004]

43 minutes ago, Beeblebrox said:

Hiya. Thanks.

Here it is again (I think you missed it the other day ;)) 

https://forums.atariage.com/topic/200118-images-generated-by-rastaconverter/?do=findComment&comment=5299136

 

 

Edited November 2, 2023 by Atari8man2004

[+Atari8man2004]

1 hour ago, Beeblebrox said:

Hiya. Thanks.

Here it is again (I think you missed it the other day ;)) 

https://forums.atariage.com/topic/200118-images-generated-by-rastaconverter/?do=findComment&comment=5299136

 

I sure miss the post I'v been in and out the the last few days. The images look great.

Edited November 2, 2023 by Atari8man2004

[Beeblebrox]

37 minutes ago, Atari8man2004 said:

I sure miss the post I'v been in and out the the last few days. The images look great.

No worries, was just being cheeky heh heh. Thanks for the pics.

[Beeblebrox]

7 hours ago, Atari8man2004 said:

Artificial intelligence 2 - Computer generated art.

 

Atari8man_artificial intelligence 2.xex 22.08 kB · 4 downloads

love this one - colours are rich, source image plays nicely with RC - good choice - nice job

 

[Beeblebrox]

10 hours ago, Atari8man2004 said:

Sifi_Lady - Mask greatly improved image.

Atari8man_Sifi_lady.xex 22.08 kB · 8 downloads

Another couple of tips.

 

With an image like this above, if you want to give RC less to process and also less to distract from the main subject, edit out any multi colour background noise from the source in an image. Obviously sometimes you want to keep details in the background so it's a trade off. Clever editing is the key.

 

I had an image I loved and converted (I did it twice) called headphonesgirl. DrPeter kindly did an attempt with it the other year with the background removed. I like both my busy background version, my version with the background removed, and also DrPeter's. Again, all about experimenting.

 

My original conversions: The first which I love despite the errors https://forums.atariage.com/topic/200118-images-generated-by-rastaconverter/?do=findComment&comment=4953847  and again after being inspired by DrPeter's version a flipped minus the background version: https://forums.atariage.com/topic/200118-images-generated-by-rastaconverter/?do=findComment&comment=4970071

 

    

 

on my mini CRT at the time:

 

AND

 

DrPeter's with no background (also flipped) which looks great:https://forums.atariage.com/topic/200118-images-generated-by-rastaconverter/?do=findComment&comment=4960248

 

***

Another little tip I've learned with masks is you actualy don't hone in on the smaller area you want to enhance. Instead you do a wider mask in a technique for some reason I saw fit to name "nibbling", as it looks like a small snial has eaten around the edges.  Don't ask me why I felt the need to name it.

 

So typically for an image with a mask you might do this  -(which is fine - but you can get really bad lines and errors in the areas of a pic that aren't masked as a result):

 

So, source image:

Typical mask you might create for it:

 

There is nothing wrong with this, however you'll likely find that outside of the masked area, using the above mask will result in a lot of horizontal lines/banding and it means the masked area looks good, but the rest of the pic suffers, typically the background if you have a central subject you are focusing on. This also gets worse if you have the mask stregth sry quite hi in RC GUI settings.

 

Instead, when I converted this particular image last year, I used what I normally use with my conversions, a "nibbled" mask... so this:

 

The mask almost covers the entire screen, which you would think is kinda pointless. However, it still helps focus the conversion and create a better result. It's about thinking how the unmasked spaces will look given you aren't prioritising them in the conversion, so they'll be susceptible to more errors.

 

You can always increase the amount of the image you nibble away at, thus reducing the mask coverage size. Or mask different areas separately. I use masks for every conversion, often nibbled ones, sometimes more common smaller masks.

 

The result is generally a much better conversion all round with less banding errors, more colours and a nicer conversion (IMHO):

 

 

Always bare in mind you can have a maximum 9 colour changes per horizontal line. Obviously more colours vertically. 

 

Again - I am not an expert and I won't pretend I fully grasp how these things work, but like many I've been converting for a good while and have  - through trial and error and converting 1000's of images and learning from others - picked up a few things.

Edited November 2, 2023 by Beeblebrox
adding more info

[+Atari8man2004]

4 hours ago, Beeblebrox said:

Another couple of tips.

 

With an image like this above, if you want to give RC less to process and also less to distract from the main subject, edit out any multi colour background noise from the source in an image. Obviously sometimes you want to keep details in the background so it's a trade off. Clever editing is the key.

 

I had an image I loved and converted (I did it twice) called headphonesgirl. DrPeter kindly did an attempt with it the other year with the background removed. I like both my busy background version, my version with the background removed, and also DrPeter's. Again, all about experimenting.

 

My original conversions: The first which I love despite the errors https://forums.atariage.com/topic/200118-images-generated-by-rastaconverter/?do=findComment&comment=4953847  and again after being inspired by DrPeter's version a flipped minus the background version: https://forums.atariage.com/topic/200118-images-generated-by-rastaconverter/?do=findComment&comment=4970071

 

    

 

on my mini CRT at the time:

 

AND

 

DrPeter's with no background (also flipped) which looks great:https://forums.atariage.com/topic/200118-images-generated-by-rastaconverter/?do=findComment&comment=4960248

 

***

Another little tip I've learned with masks is you actualy don't hone in on the smaller area you want to enhance. Instead you do a wider mask in a technique for some reason I saw fit to name "nibbling", as it looks like a small snial has eaten around the edges.  Don't ask me why I felt the need to name it.

 

So typically for an image with a mask you might do this  -(which is fine - but you can get really bad lines and errors in the areas of a pic that aren't masked as a result):

 

So, source image:

Typical mask you might create for it:

 

There is nothing wrong with this, however you'll likely find that outside of the masked area, using the above mask will result in a lot of horizontal lines/banding and it means the masked area looks good, but the rest of the pic suffers, typically the background if you have a central subject you are focusing on. This also gets worse if you have the mask stregth sry quite hi in RC GUI settings.

 

Instead, when I converted this particular image last year, I used what I normally use with my conversions, a "nibbled" mask... so this:

 

The mask almost covers the entire screen, which you would think is kinda pointless. However, it still helps focus the conversion and create a better result. It's about thinking how the unmasked spaces will look given you aren't prioritising them in the conversion, so they'll be susceptible to more errors.

 

You can always increase the amount of the image you nibble away at, thus reducing the mask coverage size. Or mask different areas separately. I use masks for every conversion, often nibbled ones, sometimes more common smaller masks.

 

The result is generally a much better conversion all round with less banding errors, more colours and a nicer conversion (IMHO):

 

 

Always bare in mind you can have a maximum 9 colour changes per horizontal line. Obviously more colours vertically. 

 

Again - I am not an expert and I won't pretend I fully grasp how these things work, but like many I've been converting for a good while and have  - through trial and error and converting 1000's of images and learning from others - picked up a few things.

I was going to ask you the other day if you can  have multi-mask parts like in the example picture.  Thanks for the info on the "nibbling" type mask and color banding.   Thanks!

[+Atari8man2004]

5 hours ago, Beeblebrox said:

love this one - colours are rich, source image plays nicely with RC - good choice - nice job

 

I did this one last month I didn't post it cause I though it was not good enough.  I put it on a slide show and somehow it grew on me  

[Beeblebrox]

43 minutes ago, Atari8man2004 said:

I was going to ask you the other day if you can  have multi-mask parts like in the example picture.  Thanks for the info on the "nibbling" type mask and color banding.   Thanks!

no worries

 

Yup, that's fine as a mask. In fact that's propably how most people use the masks. 

 

Edit: just to add - and I am not sure if it makes any difference to how the mask works - but I always reduce my masks to two colours - Black and white. I see you have grey in there.

Edited November 2, 2023 by Beeblebrox
adding more info

[Beeblebrox]

11 minutes ago, Atari8man2004 said:

I did this one last month I didn't post it cause I though it was not good enough.  I put it on a slide show and somehow it grew on me  

Most of us will convert 10 images and only post 2, discarding the others. Then weirdly you revisit those you discarded later and like them.

 

You have been converting some great images. I love the creepy clown.

 

I reckon I've converted maybe 2000 images, possibly more. A hugh amount I've discarded. Others on this thread will have been converting way more than me, and discarded just as higher a percentage.

 

Again the key is to experiment and find something that works.  But I won't go back to not using masks.

Edited November 2, 2023 by Beeblebrox

[+Atari8man2004]

AI bot grayscale.

Atari8man_AI_bot_grayscale.xex