Dark theme?

@ Rob

More to the point when can we take it out to play.

A bit of patience please! As mentioned above, I need to finish the new Theme Manager and have a front end for it and then I’ll add it to the DME Suite, so it’ll be one click away soon.

I vividly remember being a young child in the early to mid 90s and messing with the Palette options in RISC OS. I remember you could invert the colours, and that was basically like what dark theme is today.
I’m assuming that doesn’t really work anymore on modern RISC OS.

I’m about 80% sure that you can save a palette file from OS 3 and open it under OS 5.

I’m assuming that doesn’t really work anymore on modern RISC OS.

Correct. It works only with icons, furniture that uses those 8 grey scales defined in that palette, which from RISC OS 5.29 is not the case even for the buttons.

Also, simplifying a Dark Mode is just invert the 8 colours in the first half of the traditional 16 colour desktop palette is like saying we haven’t rewrite RO Kernel in C because we are just lazy. Try to invert those colours¹ and see how ugly it looks XD

¹ Yes, you still can process and produce that palette on RO 5, just through some coding (nothing too complex).

I’m about 80% sure that you can save a palette file from OS 3 and open it under OS 5.

It should work yes, but it’s easier to read and modify the Palette using some BBC BASIC code. However, for the actual results read up :)

DDT UI looks better ;)

There aren’t enough users as a whole… That train has passed many, many, many years ago and then another opportunity arised with the early Raspberry Pi and RISC OS was not ready for it, so also that train has now passed.

Using that logic then, we need to be ready for the next train. :D

Seriously, though, think how many times have we heard of the last chance for RISC OS, or that there is no future for it? Yet, here it is, open source and available on several cheap platforms, with stuff still being developed, albeit not as when Acorn dominated the British schools market, of course. Plus, I’m not alone in making a return. There’s just something fun about using RISC OS, despite it’s shortcomings.

I could list the benefits of RISC OS and RISC OS programs, so I will. There’s the small foot print of RISC OS, it’s snappy speed on Raspberry Pi boards, the modular approach instead of bloatware programs trying to do everything (think Excel), lack of unfair subscription models, and a very pleasant GUI that, whilst some feel it looks dated, nonetheless remains untouched even by MacOS in its pleasant nature when in use. !Iris has just about solved the Web browser issue, so for the end user who doesn’t care much about things like pre-emptive multitasking, as long as things work, there’s just the matter of more modern vector and bitmap graphics editors, and I certainly wouldn’t object if anyone got bored and ported MuseScore over.

I think it’s brilliant that there are dedicated people, many with brains that are surely too big to fit comfortably in their head, who have brought RISC OS this far and continue to work on it for the future. I have complete faith in them and that’s why I’m preparing to move a lot of stuff over to RISC OS. With their dedication and the stellar work they do, there’ll be a point where RISC OS will be attractive to coders who want something different and not be one of tens of thousands coding on Linux. Once that happens, the scene will be reinvigorated (remembering there doesn’t need to be hundreds of coders, or millions of users).

I have been a RISC OS user from the beginning, and I have spent (wasted?) much time on on theming, the toolsprites, the desktop, StrongED’s base mode, … etc, to my own tastes. My preferences are for minimalism, light better than dark, 3D depth over flatness, and I hate slanted directory icons. They hurt my eyes. How I wish there were more generic tools for translating spritefiles between screen modes. I have to say that since the coming of the Raspberry Pi I have seldom used any mode other than 16 million colours at 1920×1080, though on the Pinebook(Pro) I do use big mode occasionally. Still, spritefile naming is a mess: !Sprites, !Sprites11, !Sprites22, 5Sprites, 4Sprites, etc. I cannot keep up with them all. This is an aspect that cries out for reform.

May I also admit that professional graphic designers have a lot to teach us amateurs. Richard Hallas’s work and writings taught me that. The amateur tendency has been strong among RISC OS users, and has produced some lamentable results. To produce a better GUI we need at least two things: the humility to learn from people with more experience, and better tools.

The amateur tendency has been strong among RISC OS users, and has produced some lamentable results.

Definitely.

To produce a better GUI we need at least two things: the humility to learn from people with more experience, and better tools.

Do you mean better tools like graphics editors?

The amateur tendency has been strong among RISC OS users, and has produced some lamentable results.

There’s more to it than learning and better tools.
One needs something resembling talent as well.

I freely admit that my icons are complete crap. But then I have the drawing abilities of a blindfolded five year old…

Do you mean better tools like graphics editors?

Yes. Specifically for translation of sprites between modes, and tools that understand more about colour, relative weight, and the amazing psychology of human vision. I would like to see better use of perspective. Can ChatGPT help here? “Put these objects further into the background”, “elevate the source of illumination by 20%”, that sort of thing.

One needs something resembling talent as well.

I wonder about that. I think culture and society have a big role to play in visual education. Compare Danish and British popular tastes in furniture, clothing, dress. There is a world of difference. Why? I think there is a lot we do not understand about such matters.

May I also admit that professional graphic designers have a lot to teach us amateurs. Richard Hallas’s work and writings taught me that. The amateur tendency has been strong among RISC OS users, and has produced some lamentable results. To produce a better GUI we need at least two things: the humility to learn from people with more experience, and better tools.

Amen to that Gavin, couldn’t have put it any better with my basic English.

On top of the issues you’ve mentioned, now we have Transparency of the iconbar controlled by the Pinboard 2.0 configure UI… some parameters to redefine buttons colour schema in the Wimp, which is great (thanks Sprow and co for this work), but that is not back-ported, so I am still feeling the whole “painful experience”

My preferences are for minimalism, light better than dark, 3D depth over flatness, and I hate slanted directory icons.

Worry not! :) The Theme Manager already has your very own theme, so you’ll be able to enjoy RO in the way you desire. It also has traditional themes, included, for the absolute retro-lovers, RISC OS 3 theme. So there is something for everyone (I hope)

Richard Hallas’s work and writings taught me that.

Indeed, and, unfortunately, I am no way near to what Richard could do, so had to get “extremely creative” and built something that can, somewhat, mimic the process of creating congruent icon sets and for all the various formats (included the 5 and 4) and for all the themes… As both passionate for maths, you can determine the amount of work here, so there is probably no other way, unless someone is willing to take on this big task.

Yes. Specifically for translation of sprites between modes,

I wish it was that “simple”… There is more, a graphic format not supported, so most available tools out there will then need to still have an automation to convert whatever format they generate into Sprite and, again, when you multiply the effort of a single theme that may need to have all those !SpritesXY and ZSprites, even mastering whatever tool still requires a post-production process to generate the “deliverable”

So, I am open to ideas, but so far the AI is the best way to “understand” what has been done and “port it” to new themes as well as re-generating it for the new formats or for the missing ones in the icon set used to generate the analysis.

Hope this makes sense.

My limited experience and understanding of AI tells me that most of it is based on pattern matching. Whereas ordinary programming is based on Boolean logic (these two quantities are either the same or not the same) AI has to deal with multi-valued logic (the same-ish in the left hand corner). When millions of pixels and colours are involved it makes a lot more sense to use AI than !Paint. Furthermore, I think AI can be used to recognize things that can escape an untrained human eye. Apple did a lot of research into GUIs. Has any part of that work come into the public domain?

Can ChatGPT help here?

OpenAI, perhaps, but not ChatGPT. Visual stuff is more DallE2’s realm.
I asked it to create a picture of a sulphur crested cockatoo in the style of Johannes Vermeer. https://labs.openai.com/s/JqCHdE9Ar3aqpIiriKLLklji

Sometimes it can be, um, a little crazier when it doesn’t entirely understand. I mean, a chess piece king riding a skateboard at Trafalgar Square shouldn’t be too hard, right?
https://labs.openai.com/s/NsCTbkJm3aqfCGTipYuL5QHJ
The inside of a subway carriage…with cats! https://labs.openai.com/s/13Bfdrp2O1lrpfoVNR04YiA6

Sorry, got sidetracked. Cats. It happens.

furniture, clothing, dress.

For me: wooden and non-fancy, practical, black.

There is a world of difference.

Different environment, different upbringing, different cultural history.

It would be more of a surprise if things were the same.

My preferences are for minimalism, light better than dark, 3D depth over flatness, and I hate slanted directory icons.

Minimalist, dark, 3D, and luckily I’ve never owned a system with slanted icons, though I have seen them and it’s probably best if I keep my thoughts on that to myself.

Apple did a lot of research into GUIs.

They might have, but they pretty much ignored a lot of it when they decided in iOS7 to go for an ugly-as-hell flat look. And days later had to rush out an update to slightly highlight buttons and such so people would know where to prod.
Sure, skuomorphism got a bit out of hand, but there was actually a reason why buttons looked like buttons and knobs looked like knobs.

Sadly, Google played follow the leader and Android started looking flatter.

At least, rounded icons aside, nobody has arbitrarily mucked around with the UI. If I wanted crap flat icons I’d be using RISC OS 2.

I wanted crap flat icons I’d be using RISC OS 2

We were assured a graphic artist would be along to design the icons. They were not, so you were saddled with our efforts for some years.

I meant the actual Wimp_CreateIcon ones. ;)

A bit more bevelling would have helped back then. There was no need to wait soooo long for the whole of NewLook.

I am afraid:

Can ChatGPT help here?

OpenAI, perhaps, but not ChatGPT. Visual stuff is more DallE2’s realm.

No and no :(

The problem is, and believe me when I say I have tried like 6 or 7 services to date:

The general “graphic AI” (that includes also MidJourney btw) has been trained to be creative (again whatever that means in the AI world). We absolutely do not need that, because that will generate (unavoidably) icons that makes no sense to a human.

If you read again about the model I am developing, it’s not meant at all to be creative, not even a bit.

Premise: besides of the enormous effort that what follows requires (in terms of SKILLS and TIME), I am also paying with my pocket for the processing power required to train a model. I am not asking for anything to this community, I am just providing insights on how it actually works, especially in reference to some troll who is keeping trolling (now with feathers and more nonsense), clearly people don’t have any idea of how much work is going on behind the scenes here.

It processes exsisting icons and extract the common patterns, for instance the pattern that we all know well and can identify as a “File” or a “Directory”. Ok? If this is something complicated to picture, let me know and I’ll share some early successes that makes it very easy to understand what the AI actually does.

Next, the AI extracts these patterns that we all know well and transforms them into a set of declarative rules in a special “graphic language” that I have designed and that is expressed using YAML, but YAML is only the structure, the language is a real program that processes that YAML and “compiles” it back into graphic icons (this, BTW, using vector graphics, something like RO Draw). For the geeks, the compiler is written in Python 3 and that is because this way it’s easy to run it within GitHub Automations, so, when completed, we’ll be able to regenerate icons for everything that will come next.

During that compilation process it also “sucks” in (includes) other YAML files that provide a description of the colourset we wish to use for the icons drawing and a “style” one which is capable of basically saying use 16 colours, 256, 16M etc. use Alpha blending and so on and so forth.

Finally it has one last YAML that foundamentally contains a static list of ALL the icons we want to generate. That has reference to the shapes and colours in a human readable form (like everything else).

So, you set which style and colourset in a parameters and run the build and it starts to generate all the required icons in the style and colourset specified.

That is what we want. Given, let say, we may decide to sligthely change a tint (liek Rick suggested in another post on another thread), then he’ll need to start a full regeneration to make sure everything matches and to select which icon style he wish to use for his desktop theme.

Obviously, nothing of this is preventing a human to design an entirely new icon set, but that is a different story (which so far, no one has offered to help with).

Apple did a lot of research into GUIs.

Indeed and there are reseaches that have been reaching the public, some of which are about rounded corners and others are about hiding options all to improve the perception that both the OS and the Apps are easy to use. That doesn’t actually makes them easier to use, it only gives a perception of it.

There are also other historical book/researches like Dont Make Me Think, which focuses more on the age of Internet Apps, but most of the principles are applicable to traditional GUIs too.

Finally, there is the problem at hand, the so called “de facto standard” on the matter of interacting with Applications and graphics Desktop, which in this day and age DOES have rules. RUles that were not written at the time RISC OS was dersigned, so the question there is: SHould it adapt or not?

My DME answer to this is: Both. You can use the DME to adapt it to modernised workflows or just keep it the way it has always been.

Compare Danish and British popular tastes in furniture, clothing, dress. There is a world of difference. Why? I think there is a lot we do not understand about such matters.

That’s an interesting point. I find traditional Japanese design to be very appealing, and very calming. I’m not surprised it’s different than traditional British design, especially considering the periods of isolation, but I think it’s notable for how right and balanced it all seems, especially the way it fits in with the natural environment around it. Then again, they get a lot right, like not getting rid of old stuff just because there’s something new.

https://www.youtube.com/watch?v=zWJZFQHklBg

If it does the job…

Sorry if this is right in front of me, but is the aim of Theme Manager to actually generate icons and then manage them, or is it to allow someone to create a theme themselves and then it manages it? I’ve got a bit lost as it goes into AI talk etc.

Theme Manager to actually generate icons and then manage them, or is it to allow someone to create a theme themselves and then it manages it?

Both.

The Theme Manager protocol is compatible with a tool I have linked in another thread, this is to make life easier for the user who whish to design their own Theme. Either from the ground up or just the Windows funrinutes and colour schema.

At the same time, for the pre-generated themes (which are a lot already) or for users that do not wish to redraw ALL the existsing icons into their own style, I need an AI to help me in this very complex process of pattern analysis and redrawing all the icons you see in front of your eyes on RISC OS (at least the OS alone) so that the pre-generated themes do not appears “disfunctional” in their aspect or whatever is the graphic designers language there.

The “icons compiler” is almost completed, the AI is still on training, this normally takes a lot of time and I am also ensuring it can process icons from different OSes, to make it possible for users to also generate modern themes.

The inside of a subway carriage…with cats! https://labs.openai.com/s/13Bfdrp2O1lrpfoVNR04YiA6

That’s reassuringly bad. I read that AI (I think the article was on about ChatGTP specifically) gets around half of what it is asked wrong, but it obviously doesn’t know it is wrong and presents it as fact, similarly to how that AI has not realised it’s made a dog’s breakfast of the cats in the subway train. It’s kind of a weird concept that it churns out creative stuff but has not idea if it’s crap or good.

Paolo: Remarkably, I think I actually understood that. Interesting use of AI, too. Admittedly I’m usually against AI but always open to being convinced it can be used in a good way to help us.

I’m sure I read in these forums that icons should ideally be rendered as vectors. That is to say, it would make resizing the desktop etc etc a lot easier regarding icons. Is that an aim of the software?

I’m sure I read in these forums that icons should ideally be rendered as vectors.

Yes, this is the “future”, right now, we have already a lot of problems to deal with as things are. But, as an example, if you look at my Launchpad, you can see it already creates Apps icons as 2 seprarate icons (the sprite and the name), this, for example, already allow to scale up an icon for people with very large screen resolution. But I haven’t enabled such a feature yet, because there are plenty more to enable first and have people helping to test it (well) to test those.

Never a dull day… (ermmm night!)

Is that based around the !Draw format?