I'm not sure that there should be much distinction between "clothed" and "naked" from the perspective of description generation. If I'm wearing a brightly patterned shirt, is that any more or less interesting than if I'm covered in tattoos or scars? And isn't the latter really just another layer that can be hidden by worn equipment?


Interesting idea, but won't that end up with a sentence for each item? And how would it deal with something like a cloak (conceals the fact that your back is exposed, regardless of what you're wearing on your chest), a breastplate over a shirt (the former on top, but the latter still visible on your neck and arms), a robe (covers the arms and legs as well as the torso), etc?


The full helm would hide both your hair and your facial features. However you could also wear a hat which only hides your hair, or a mask which only hides your facial features - and of course you could wear both a hat and a mask. Likewise, even if you're wearing a full helm, you might have the visor pulled up, and/or have the hood of your cloak pulled over the outside.

How would your approach generate sentences based on multiple items?

I seem to recall someone bringing this up before, but I don't feel like rereading the thread (three whole pages?!), but a "shininess" factor would probably work well here. Each item would have a value that indicates how attention-grabbing that item is. Paired up with multiple ways of generating descriptions, you could have something along the lines of the traditional long description showing "A male human wearing a suit of flaming plate mail is here" while various other "look" commands would display sentences based on the most interesting features (which might vary by character). Bare skin would have fairly low "shininess," but scars and tattoos would raise it.

If a given piece of equipment doesn't have 100% cover (opacity?), throw some of the the layer(s) below into the equation and mention them too, giving you something along the lines of, "A tattoo of a fire breathing dragon is etched into Bob's chest, partially covered by a glittering mythril chain shirt." It might be better to order the items in this construction from low to high shininess (Bob's skin is partially visible beneath his old, rusty chain shirt). A quick glance would display only the X layers with the highest calculated "shininess," while a deliberate study would perhaps display (most?) every slot. To avoid randomness, some sort of awareness mechanic could be worked in to not display items with a "shininess" below the highest displayed value (so a person holding a blinding mote of the sun's bland iron ring would go unnoticed without an observer having a decent awareness score).

A little from category "On Topic" and a little from category "Off Topic," but I felt like blathering instead of doing real work. :biggrin:

One way I would aid the % coverage, is by taking away the ability to write paragraphed descriptions. You could menu-ize traits and create the description from what the player selected. Then, you can set equipment coverage %ages to remove those descriptive sections.

For something like a helmet with visor, is saw on one mud, a mask capability where if the player raised a hood or donned a mask, their long description changed to a male human wearing a pale white porcelain mask. Or if it was a hood, a male elf wearing a hooded flowing cloak. Now, you can take that and just remove the specific covered descriptors if the hood was raised, and return them if the hood is lowered.

You could also take all wear areas and have the player describe each bare location, and then with coverage, remove the wear location description if the item covered it.

Late reply I'm afraid, as I only just noticed the response - so much activity on MudBytes lately that the thread was pushed off before I saw there was a reply.


Yes, that's the plan - to generate the entire description, so viewers can only see what they should be able to see. But it's not as simple as just removing sections, because a single sentence may be constructed from multiple traits, and different combinations of trait may require the sentence to be written completely differently. For example with the head/face situation alone you've got the following possible combinations:

1) A single item covers both head and face. Eg:

A pair of steely grey eyes gaze out through the slit in the closed visor of his full helm, which fully conceals the rest of his features. He is wearing a breastplate over his chainmail shirt, and a pair of runic bracers are strapped to his forearms. He is wearing a codpiece over his chainmail leggings, with greaves strapped over his shins. He is over twenty-six and a half feet tall, and moves with cat-like grace.

2) One item covers your head, another covers your face. Eg:

A pair of deep blue eyes gaze out through the eye holes in his mask, which conceals his facial features, and the hood of his cloak is drawn up over the rest of his head. He is wearing a shirt, the long sleeves coming down to his wrists, and a pair of runic bracers are strapped to his forearms. He is wearing a codpiece over his trousers, which are tucked into his boots, with greaves strapped over his shins. He is over eight and a half feet tall, and moves with the grace of a dancer.

3) One item covers your head, but your face is uncovered. Eg:

He is hideously ugly, with stoney grey skin and glimmering opal eyes, and the hood of his cloak is drawn up over his head. He is wearing a breastplate over his ringmail shirt, and a pair of battle bracers are strapped to his forearms. He is wearing a codpiece over his robe, which hangs down to his boots. He is about six feet tall, and seems quite agile on his feet.

4) Your head is uncovered, but an item covers your face. Eg:

He has long wet auburn hair which reaches down to the middle of his bare back, and his light brown eyes gaze out through the eye holes in his ripped-off face. His dark brown naked body ripples with muscles, with a thick neck, broad shoulders and sinewy arms and legs. He is over five and a half feet tall, and moves with cat-like grace.

5) Both head and face are uncovered. Eg:

She is quite cute, with lightly tanned skin, steely grey eyes and long soft auburn hair which reaches her bare shoulders. Her naked upper body is quite muscular, with small breasts, a flat stomach and a narrow waist. She is wearing a pair of trousers tucked into her boots. She is about six feet tall, and moves with the grace of a dancer.

Furthermore, (4) and (5) could also have a nonblocking item worn on the head (i.e., something that should be added to - or even modify - the head description, rather than simply hiding it), eg:

He is reasonably attractive, with pale white skin, glowing red eyes and long curly black hair which reaches his bare shoulders, held away from his face with a bandanna. Thin lines of dark runes cover his cheeks and naked upper body, tracing the muscles of his chest, coiling around his bare arms and twisting across his flat stomach. He is wearing a pair of trousers tucked into his boots, with greaves strapped over his shins. He is nearly six and a half feet tall, and seems quite agile on his feet.

Then you start getting into items that also hide the eyes (eg dark shades, mirrored visors, etc), helmets that also cover the neck, and so on. And that's just from the neck up, which is relatively self-contained - when you start looking at the rest of the body, which can wear all sorts of clothes that cover all sorts of combinations of body parts, it gets a lot more complicated.



Do you mean 'long description' (when you look at someone - i.e., like the examples I've given above) or 'short description' (what you see instead of a name, usually no more than three or four words)? If you mean the latter then that's pretty simple, as the descriptions are so short you can just provide a few alternatives. Concealing an entire long description just because you've raised your hood or donned a mask isn't something I'd like to do though.


Perhaps, but personally I prefer generated descriptions. Less error-prone, less abusable, and more consistant. Plus I sometimes like to tie in human form descriptive traits with shapechanged forms (eg a blonde werewolf has white fur in wolf form).

When I was toying around with Smaug, I had an issue that's semi-related to this. Namely, a hooded cloak can cover the body AND head, but doesn't have to. A shirt with sleeves covers the chest, arms, and possibly wrists. I wanted it to show up with stock smaug descriptions such that you could see these things.

e.g.,
<body, arms, and wrists> a shirt

And so on.

What I decided to do was give items a list of possible wear locations instead of a single slot. Then I figured I could give some of those toggle-able states. So if you have a helm with a visor, the helm's wear location is "head", but it also has a toggle-able wear location of "face", so if you "lower visor", it now covers your face and head. Then you'd have to provide a bit of checking so that you can't lower a visor down over a mask (where it wouldn't work if the visor was already lowered) and people could adjust their eq while they're wearing it. Such as roll up sleeves to wear bulky arm-guards, that sort of thing.

With the mechanics worked out, I'd just have to edit the look code a bit. Mainly I'd just daisychain the multiple locations with "and" and commas.

How you'd display these sorts of things as dynamic text, I don't know, I wouldn't really know wear to begin, but if you can display it as stock eq with multiple slots, then you can display them as dynamic descriptions, once you've worked out the kinks.

This just popped into my head while I was typing this, but you could also give sublocations keywords. A "shirt" covers the body, a shirt's sleeves cover the arms. A helmet's visor covers the face, and so on. If you can come up with a name for each sublocation, you could treat them as individual items, where "helmet's visor" is the name of the "mask" he happens to be wearing.

I already support equipment covering multiple locations and/or layers, but if anything that actually makes it harder to generate descriptions, because there are more possibilities to take into account.

What I mean is, if you can already display a shirt as "<worn on body, arms, and wrists>" you have the beginnings needed for a dynamic description.

"He is wearing a shirt on his body, arms, and wrists."

I think I'd be tempted to prioritize wear locations. I don't pay much attention, for example, to people's shoes. If I notice them at all, I notice what they are wearing on their face/head, their upper torso, and their legs. Other locations should probably be relegated a more in-depth "look" command, like examine, barring "shininess". Of those four possibilities, I'd say the head/face should be a separate line from the upper/lower body.

Since it's easier for me to write python code than to think, I wrote some python code to generate random dynamic descriptions. I had to ensure that each eq string was a singular phrase since I didn't feel like tracking singulars and plurals. I get output like:

tonitrus@wander ~ $ python3 dynamic_descriptions.py
Test appears to be a dwarven female, approximately 43 years old.
She has blue eyes, grey hair, and light skin.
She is wearing a bronze helmet with a steel visor on her head, a suit of chainmail on her torso,  and a pair of pants on her legs.
tonitrus@wander ~ $ python3 dynamic_descriptions.py
Test appears to be a human male, approximately 37 years old.
He has blue eyes, red hair, and reddish skin.
He is wearing a steel helmet with a steel visor on his head, a steel breastplate on his torso,  and a pair of leg-guards on his legs.
tonitrus@wander ~ $ python3 dynamic_descriptions.py
Test appears to be a human female, approximately 22 years old.
She has brown eyes, grey hair, and dark skin.
She is wearing a set of steel leg-guards on her legs.
tonitrus@wander ~ $ python3 dynamic_descriptions.py
Test appears to be a dwarven female, approximately 43 years old.
She has hazel eyes, grey hair, and light skin.
She is wearing a steel helmet with a mithril visor on her head.
tonitrus@wander ~ $ python3 dynamic_descriptions.py
Test appears to be a dwarven male, approximately 27 years old.
He has blue eyes, grey hair, and light skin.
He is wearing a steel helmet with a bronze visor on his head and a pair of pants on his legs.
tonitrus@wander ~ $ python3 dynamic_descriptions.py
Test appears to be a gnomeish male, approximately 37 years old.
He has blue eyes, red hair, and light skin.
He is wearing a steel breastplate on his torso and a set of steel leg-guards on his legs.
tonitrus@wander ~ $ python3 dynamic_descriptions.py
Test appears to be a elven male, approximately 36 years old.
He has blue eyes, grey hair, and reddish skin.
He is wearing a steel helmet with a mithril visor on his head and an unremarkable tunic on his torso.
tonitrus@wander ~ $ python3 dynamic_descriptions.py
Test appears to be a elven female, approximately 20 years old.
She has hazel eyes, red hair, and light skin.
She is wearing a bronze helmet with a bronze visor on her head, a steel breastplate on her torso,  and a pair of leg-guards on her legs.

Now admittedly, these aren't pretty. There's some repetition, I should probably use sentences that track grammar instead of stock strings and pull out repeats like "a bronze helmet with a bronze visor" so that it reads "a bronze helmet with a visor", but it also doesn't look terrible, and I was able to hack it together in a few hours with python.

I suspect that the reason that it looks as passable as it does is that I'm handling only a few locations and only have one "sublocation" (the helmet's visor). My suspicion is that if I add sleeves and such it will start to look unwieldy, and will look worse and worse with each potential field of text I add. I'll let you know as I get around to doing that.

If I'm right, then the secret will be just ignoring as many fields as you can, since trying to form readable paragraphs for such things is probably going to require much more effort than it's worth.

I think it's probably also important to have separate lines of interest. I went with general information/general description/equipment for that reason. It probably would have worked better if I went with something like "She is wearing a bronze helmet on her head, with a bronze visor covering her face" instead of trying to daisy-chain helmet and visor into one text, but I'm tired of messing with it for now.

In the category of my Weirder Ideas™, another thing you could do is have multiple optional displays for each line of text and allow the player to select between them. I.e.,

[1] a bronze helmet with a bronze visor
[2] a bronze helmet with a visor
[3] a bronze helmet, with a visor covering your face

That idea isn't really ideal either, but it might give you a leg up on implementing dynamic descriptions without having worked out every arrangement that results in grammatical ugliness.

One of the ideas I tossed in a while back is to not bother mentioning things that aren't remarkable. I'd also suggest not mentioning where the item is being worn if there's only one logical place to wear it. The above example would become.


Note that this assumes the visor is up, and that his hair is long enough to be visible despite the helmet. If the visor were down, you'd skip the eyes (since you can't see them), and if his hair were short, you'd skip that (since you wouldn't see it either). There's no reason to mention he's wearing it on his head… where else would he be wearing it? Since the tunic isn't remarkable, why bother showing it? Just being absent from the description does not imply he's running around bare-chested… in fact, that WOULD be notable.

The idea is, you want the text to read just like you were encountering the character in a novel. Overly verbose is just as bad as overly terse. Unless you're reading Dickens, perhaps. :)

But my point is not everyone will be wearing a shirt - you'd still have to cater to the people who are just wearing a vest (body only), bracers (arms only) and bracelets (wrists only), just as a single-location mud would do. So you've actually got more possibilities to take into account.

In addition, because I use layers, I also need to deal with things like breastplate and bracers worn over the shirt.


Which IMO is a good solution in theory, but in practice could result in some strange scenarios like:

You tug an unremarkable tunic vest onto your body.

You pull your legs into a pair of unremarkable leggings and button them up.

Bubba looks at you.

Bubba says, 'Dude, you're naked!'.

Perhaps it's not quite so bad if you don't generate naked descriptions - but I explicitly want to include naked descriptions (something which is of particular interest when looking at monstrously mutated people).

However I can also imagine it being difficult to generate descriptions based on layered equipment when some of those items aren't being mentioned. For example supposing I'm wearing an unremarkable jacket over my superman t-shirt - the jacket shouldn't be mentioned as it's unremarkable, but that doesn't change the fact that (when closed) it's hiding the logo on my superman t-shirt. Should that make my t-shirt look unremarkable as well? Or does the t-shirt make my breastplate remarkable?


Generally agreed, although for some items it might not be clear where they're supposed to be worn. Also, when the sentences get quite long it can sometimes be easier to mention some locations for the sake of clarity - for example:

"You are wearing a codpiece over your trousers, which are tucked into your boots, with greaves strapped over your shins."

How would you write that without mentioning shins? You could just end with "strapped over the top", but to me that might read as if the greaves were strapped over the boots or perhaps even the codpiece. You could also say "You are wearing a codpiece and greaves over your trousers…", but that reads a bit strangely to me as well.


Indeed, and if you want hair and eyes displayed as part of the same description it can start getting quite tricky.

I don't think the unremarkable idea would make anyone naked. What Quix has mentioned is to make certain pieces more remarkable than others, and only display the most remarkable. So if you were wearing a bunch of boring crap, the most remarkable of that boring crap would show, with certain wear locations being weighted to be more special than others.

So instead of Bubba seeing a naked guy, he would see boring crap the guy was wearing. If the guy was naked, Bubba would then see a naked guy. You could certainly make a few wear locations very remarkable if naked, namely pants and shirt.

With this setup you could pick the amount of locations you want to describe and keep it smaller while still pretty dynamic. I think you'd still have the problems of making the sentences fit together nicely however it would dramatically reduce the size of the problem.

It would if all of their clothing was unremarkable, but they were wearing a load of remarkable rings, necklaces, bracelets, etc.

As I said before, perhaps it's not quite so bad if you don't generate naked descriptions - but that's something I do want to do, for a number of reasons.

What the proposal has been is not simply avoid showing "unremarkable" eq but instead to compare the equipment and show the most remarkable over the boring, with certain wear locations having more weight. If the character was wearing ONLY boring eq, you would show the boring eq locations that had the most weight. Head and Torso areas come to mind.

Naked eq slots could be weighted to have a very high priority, or coolness factor IE: No pants. While not having shoes wouldn't be nearly as big of a deal.

So a good example is this. Let's pretend you decided to only show a maximum of 4 eq slots.

Bob the warrior is wearing ALL boring eq. You would see what he wears on his face, head, torso and arms or whatever other areas you find most important.

If Bob is wearing all boring EQ, except really nice shoes, you would focus on his shoes and then show the rest of the boring things. Showing face, head, torso, and shoes. <— Since arms are not a huge priority slot by default, they are taken over by special shoes.

If Bob has a lot of really cool eq on his fingers, feet, head, and face but is not wearing any pants you would have weighted that naked eq slot to be the top priority, and this would likely cause the least noticeable location to be kicked off the show list. I think that location would be fingers. So the list would be head, face, naked lower body, and feet.

So the problem of characters wearing eq looking naked is not a valid problem. It simply is incorrect with what has been proposed.

Indeed. Naked is more interesting than average stuff. :)

There's also a way around the newbie wearing all boring gear, just present a generic description until they wear something interesting enough to stand out.



Bubba sees:
You see a blonde human man, wearing typical cloth garb.

If he removed his pants:
You see a blonde human man, with no pants!

If he wielded a longsword:
You see a blonde human man, with no pants, wielding a steel longsword!

If he put on a pair of nicely fitted jeans:
You see a blonde human man, wearing flared blue jeans and wielding a steel longsword.

Hmmmmm….


I guess that depends on how much you are willing to assume about your audience's knowledge of your gear. The fact that they're greaves does imply they are shin armour, and so it's not necessary to state it. It might be more awkward doing it differently though.

You are wearing a codpiece and a set of greaves over your trousers, which are tucked into your boots.
You are wearing a pair of trousers, tucked into your boots, with a codpiece and greaves on top.

I think the key for this kind of presentation is to derive a metric for judging the complexity of finished output, and then have your constructor try multiple paths to see which ordering and form is best. In that respect, using the same grammar rules an automated grammar checker does might be a good starting point.

In the end I decided to combine the belt with sentence 3, and create two new sentences for wielded and sheathed weapons respectively, with the former describing the weapon in full and also mentioning gloves. These two new sentences are separated from the main description by a blank line, inspired by Gemstone (see post #36). I've not bothered with necklaces or rings.

Thus it's now:

Sentence 1: Appearance (or face armour) and hair (or head armour).
Sentence 2: Body, arms and forearms equipment (or naked upper body).
Sentence 3: Waist, groin, legs, shins and feet equipment (or naked lower body).
Sentence 4: Describe height and stance.
Sentence 5: (new paragraph) Listed held weapons and mention hand-worn armour.
Sentence 6: (new paragraph) List sheathed weapons.

Note that sentences 2 and 3 may be combined if they're short enough, much as I described in my earlier post - I'm still not that happy with it, but I've not found another approach I like, so I'm going to stick with it.

The mud defines "naked upper body" as wearing nothing on the body, and "naked lower body" as wearing nothing on the groin. If you are wearing nothing on either body or groin then you are "completely naked". Naked descriptions have three possibilities:

1. If you are classified as "naked" by the above definition, but you're still wearing something on your arms/forearms/hands (upper or completely naked) or waist/shins/legs/feet (lower or completely naked) then the sentence will describe you as being "naked other than…"

2. Otherwise, if you have tattoos or other markings, these will be used to create an appropriate description of your naked body.

3. Otherwise, your Brawn stat will be used to create an appropriate description of your naked body.

It would be better if 2 and 3 were combined. Maybe I'll look into doing this at some point, but it's not that urgent.

Combining sheathed weapons (in sentence 6) is still a little iffy. But this is a separate paragraph at least, so it's easier to tweak and play around with.

I've also tried extending the approach to non-human forms, including "mist form" (following some suggestions from Tonitrus). Pretty awkward to describe 'a vaguely humanoid mist' that may or may not be wearing partial or full armour, but the results are acceptable IMO.

Extending this to demons, dragons and draconians is going to be messier, and I don't think a generic solution is going to work - I fear they may each need to be handled separately. I don't consider these quite so urgent though, as the main reason for including equipment in human descriptions is to flesh them out a bit, and the other forms already have fairly extensive descriptions.

More difficult, perhaps, is expanding this to forms that don't use equipment, as a generic "naked" description just isn't going to do the job. A wolf can wear a collar, but that alone isn't enough to produce a decent description of any reasonable length. Describing "a cloud of bats" is even trickier still, and I think it might be better to approach it with a different outlook altogether. These should probably be approached in the same way as dragons, draconians and demons - the problem is they're nowhere near as customisable, so it's hard to think up a lengthy description that doesn't look obviously padded.

All in all, it's a lot of effort for something that's purely cosmetic, but it's kind of fun and I think it adds a lot of flavour.

I'm currently working on something similar. Not descriptions, as I like those hand written and feel they add to the roleplay environment. Instead, long descriptions are generated based on equipment and player status.

<player> is <posed> here, <wielding/holding/with a sheathed> <weapon/item type>, wearing <jewerly items in a string>.

In my old mud I used a recognition/introduction system, and generated short-descriptions which included held items (although nothing else). It was actually quite useful though, as you might see (for example) a "tall dark-haired man holding a pair of knives" walk into the room, and you'd know that he was ready for a fight.

I sometimes feel tempted to add a recognition system to my current mud as well, but honestly I just don't think it would fit that style of game.

My suggestion is not very well thought out, or researched but I got a gut feeling that you might want to try something like what would be done in creating a language or developing your own grammar rules. Because that is really what you are doing. But it seems as if you are trying to shoe horn defined descriptions into your output then to gracefully build your own grammar sets based upon the equipment that is being worn. I wish I could come up with an example right now, but I am drawing a blank.

I played a game called Realms of Kaos (apologize if this is super necro of a thread, but it was linked to in a recent thread) where character descriptions were dynamically generated based on gender, race, class, stats, equipment, and divinity(religion/alignment type trait).

Assassin at max level with high strength, high agility, low charisma and low intelligence.

He is a male saurian slayer. He looks incredibly strong. His movements are quick and agile. He is rather average looking. He wields an Acheron's Knife and is protected by his Blood Fiend Skins. He is a member of the clan 'Kaos Hitlist Service'. He is a follower of Targos.

Cleric at max level with low strength, high agility, high intelligence, high charisma.

She is a female gnome priest. She is slim and slightly built. Her movements are quick and agile. Her eyes shine with intelligence. She is extremely attractive. She wields a Hammer of Justice and is protected by her Titanium Platemale. She is a captain in the clan 'Shadow Society'. She is a follower of Alustria.

Not all stats are indicated in appearance (like endurance, wisdom, etc.) and neither is all equipment. Helms and shields aren't shown by default, but typically your weapon and armour are. Hope this helps somehow.

Ouch, Realms of Kaos is even more clunky than the Gemstone example in post #36.

Still, it's interesting to see examples from other games. I'm surprised more muds haven't bothered, even if only to replace the "You see nothing special" that they typically display when you look at someone.

Realms of Kaos is a fantastically addictive gaming experience that is borderline irreplaceable. In fact, I wrote my own version of the same game in C and Lua; new client, new server, database support, with classes, skills, races, rooms, items, and NPC's all scripted and able to be modified on the fly.

Unfortunately, I can't let people play it because I don't have permissions for any of the art, music, or content, nor can I get the current Administrators to adopt it because they don't really exist. It would make their lives much, much easier, and give the game a reasonable chance of continuing to live and grow in the future.

It was a fun project though, I learned a lot that is applicable to MUDs. Just wasn't stuck in telnet.

Edit: @Kavir: Realms of Kaos has images though :D you're right, still… very clunky.