Thats how we use it at the moment, we check for available images and then pick one of the options (but first we pick general category assuming there is always a pic for it or use default text/pic instead).
Still not what I mean.
I want to avoid checking for what's available. I want to pick a random image that is tagged as sex image and then parse the sex related tags to construct the text and the resulting stat increases.
No checking. No categories. No defaults. No static choice between the available options. Take any sex image, read the tags, construct the text, add relevant stats.
Note that logic in jobs is not very flexible so changing it for tags significantly might mean a complete rewrite and that's not likely to happen any time soon unless Thewlis wants to take a stab at it.
For now, that's fine. I only want to apply the concept to sex images, to see how it turns out. But if it turns out well, a similar concept could be applied for jobs at a later point in time.
Long €dit:
I had some time on my hands today, so I decided to make a few more points so you will (hopefully) never repeat this statement again:
It's prolly best to keep current loading system, just expand on it a bit and create a tagger that can write to files instead of JSON.
You really think so? Alright then, here's why not.
Pinkutako and DarkTI have already realized it: The sex tags currently are a mess and need a better format. If you don't believe me, here is a simple example.
Imagine a situation with more than two participants in the scene. For example, an image with the tags
group,
two girls,
do dildo,
dildo pussy,
dildo anal. Read the tags carefully. Notice something? You have no idea what is actually going on in the scene. The tags do not define who is doing what to whom. This problem appears for every group image and propably for several other situations as well.
You think this is not a problem? It is. Why? Because you want detailed descriptions. That is the whole point of your system. But you can't provide detailed descriptions if the tags don't indicate what's going on in the scene. You could argue that you can always use more general descriptions as a backup plan, but that's
exactely what you want to avoid with your tagging approach, so please don't bring it up - you'd be shooting in your own foot.
There is a second argument that you could bring up: Why use group images? If you're honestly asking yourself that question right now, allow me to quote Xela:
LoL You tagged 15658 images and typed this... At least Dark/CW'll have a laugh 
Read the context if you aren't familiar with it. Xela told me that prostitution images in which the prostitute isn't taking the action should be allowed because there are a lot of images with that kind of situation. While I don't necessarily agree with the relevance of that argument, it is a valid point. So using passive prostitution images is fine, but group sex is not? Not cool.
That should make it clear that an improved tag format is needed (or at least very advisable if you want to keep up with your own standards and statements). Then the next question would be: Why do I want to change the way parsing works just because the tag format gets adjusted?
To put it simple: Because the new format makes parsing the tags a lot more complex. There is a context now, which means that you'd have to consider a ton of things when looking for an appropriate image. You'd have to include and exclude so many things that I definitely wouldn't want to write the code for it.
Now the problems should be clear. But what is the alternative, the solution?
- Store images per character or don't store them at all. There's no point in having a tag database for all characters, the effort for sorting out a single one is too high.
This will reduce the amount of images that need to be checked significantly. - Filter all unwanted tags before choosing a sex image (for example by mood, amount of customers, etc.).
This will avoid having potentially unfitting images. - Pick one of the images randomly.
This is a very low performance operation and will guarantee that every image is statistically equal. - Count the amount of customers and actions and choose an appropriate frame for the text description.
You could also just concatenate the different descriptions, but that would be boring. - Insert the customers and actions into the description frame.
Given that there are multiple frames and multiple descriptions for the same situation (chosen randomly), this will guarantee a lot of diversity.
And that is better because...?
- Faster, more efficient parsing
- Less wasted storage space
- Easy to understand, fully procedural lookup algorithm
- More intuitive tagging
- Equal usage of every sex image
- More diverse text descriptions
Have fun!
