itch.io

If you can specify what kind of scenario you exactly want (Overview, goals, locations, entities(with pictures preferably)) I can fashion something, but alas due to AI context and Formamorph's not a game engine software limitations, replicating of zxc's style of having multiple bad ends and various specialized encounters in a specific descriptive and visual format (without changing pictures) is hard.

Never seen other people making too much corruption/latex stuff.
I've made a whole ton of worlds just using the themes corruption, transformation, slime, futanari, etc (but not the bondage-esque nature of ZXC's works). I could never figure out exactly the right formula for describing the exact kind and process of transformation being presented. If I could get that, making latex scenarios would be much easier, but I haven't found it yet.

Most of my scenarios are based off of whatever images or comics I find interesting, but not really games, however I have played most of the RPG Maker-based tf and tg games including Latex Dungeon.
I haven't made any scenarios directly based off of any of the RPG Maker or zxc games, nor even any "Go through X, find Y, defeat Z" scenarios, but I've drawn lots of inspiration from their work.

I don't really have much motive to make anything really much like them(Unless I get a really great template and descriptors for the AI which make it game-like but playable enough), but I do have scenarios you can edit/alter to your fitting that probably have close enough base entities and formatting that would make it much easier to actually come to fruition. Check out these for some inspiration maybe in a good order:

1. The whole of the slime_mall_world scenarios, 01400, and the individual 014XX worlds.
2. 19004. rubber_factory_world, I tried for a while with this one but never got very far.
3. 2600X. trap_cursed_dungeon_world, almost a close scenario but not really. Not enough depravity.

I dunno, I never really focused on the game side of things, and Formamorph is somewhat limited when making more gamey mechanics like encounters, multiple enemies, etc. You can find all the rest of the misc world I've made here: GDrive.

Yeah, heavily dependant on the AI unfortunately.
Tested a bunch of random things, still iffy. I'd had to detail everything in my responses (like usual), and it goes off randomly less compared to default (yours), just not much. 🤷‍♀️ Lack of vore params I guess, especially unbirth, very hard to get it to work without handholding everything.

Honestly, I found asterisks stopped being a big impact on rule following (At least for DeepSeek V3 0324), but other than their (tiny) bulk, there aren't many improvements for conciseness. Ran your Infernatus world' system prompts through gpt and my templater, and the important rules sizes char count only increased (other tweaks could be made).

Best result:

- Swallowed prey cannot see or physically interact with anything outside their predator. They can only perceive and interact with the internal environment of the predator (e.g., stomach walls).
- Predators cannot see or directly touch the prey once fully swallowed, though they may feel their weight, movements, or squirming.
- Predator vore rules apply to the player

Went back through some worlds, wanted to edit some of the images I used, but couldn't find where the original images were (Probably deleted from Downloads, oops). Anyways, images are saved as 'Base64' strings, which can be decoded into images again, easily, just json editing.

Here's the process to take images out of a world (Works on any downloaded world, helpful if you've lost the original images):
1. Download the world somewhere onto your computer.
2. Access the JSON file directly (just open it), helpful to use software like Notepad++ but Notepad or any text editor works the same.
3. In the file, under;
     "worldOverview" look for "thumbnail": and copy the quoted text
     "locations" look for "backgroundImage": and copy the quoted text
     "entities" look for "image": and copy the quoted text

4. Take this data to a site like Base64 Guru, and simply paste it there. Be sure to remove extraneous bits (quotation marks,'"image":', or anything else).
5. Simply download and use.

Additionally if you want to remove images (readability, file size, idk), you can replace the the entire data with 'null'. Don't forget the comma. 

Dictionary is a game changer, and all the updates/tweaks I was making to my decent worlds now are inferior compared to anything I could be making 😿.
Lmao, anyways thanks for the great updates.
Not putting my works up into the online share space (too much spam, maybe in the future, better works, etc). Keeping em here -> GDrive

^ 

You can change GDrive setting under 'triple dot', 'Share', 'General Access', and set who can access to "Anyone with the link", as either 'Viewer' or 'Commenter' (Editor can erase files). 

Testing a new prompt: Test NSFW Prompt 👌
Feedback is appreciated, check out the other worlds I've created, and extra saved stuff: GDrive Warehouse ❤️

Yes, I've been experimenting with lots of different stuff, that one's based off of The Exile of Aphrodisia... And there's tons of unique scenarios, so many that I got burnt out from taking encounter text and making  simple but faithful entities and transformation. Dunno, 
check out The Exile of Aphrodisia (It is free).

Thank you for all your dedication to this great application! ❤️

Take care of yourself, fun software can and should always come later than life (as shitty as life is). ✌️💜

Congrats on open-source!

I occasionally use either 'turn', 'step', or 'moves', but if consistent enough you can use 'minute', 'hour', or even 'seconds', so long as all prompts have a specific time included and context is short enough. You might also need to include a time for how long everything takes when you address the AI, ie: "For the next X minutes" or "In Y hours" or "After Z turns". ❤️

Cool! Very focused, and simple. ❤️

Reasonably impossible to fit in a 8k context limit without severely affecting the final output quality. Games do not translate well into worlds, unless the game and it's actors are already ambiguous/generic and of a well-known nature (more for AI to draw on from it's own data).

See my other reply but you can setup an OpenRouter acc or fine-tune the Endpoint and System Prompts under settings (do turn off Single Paragraph Event if you want to see actual changes lol), there's a bunch of better posts covering AI, posted by much more AI knowledgeable people.

Lol.

Tbh, the base Formamorph AI model can actually go beyond 2000 default Max Memory under Endpoint settings(found while in a world (changes maybe sometime soon)). For starters, try increasing the Max Memory to 4096 and see how the AI behaves. 

Link to official guide: Quick Setup Guide: (Free!) OpenRouter Setup
If it still is as erratic with the details, try setting up an (free) OpenRouter account and starting with nice AI like Meta's Llama, under the free filters. There's already documentation/methodology/a guide on how to connect your own API, but just paste your (newly created) API key into the Endpoint API Token space in settings and set your Endpoint URL to 'https://openrouter.ai/api/v1/chat/completions', and copy the name of the AI into the Model Name spot. So 'meta-llama/llama-3.3-70b-instruct:free', or whichever else work. OpenRouter does work on mobile devices (Formamorph's UI is atrocious on phone but very usable on tablet/ipad.).

With this, you area able to experiment and peruse the free models (there are... better ones with much higher context (>8K context tokens) and/or personality), but LLama is a fun baseline.

AFAIK, the default Formamorph AI is locally hosted, and therefore probably has server-side nsfw prompting built in(not 100% certain), so that's what seems to be prompting all the erratic nsfw content. When you use OpenRouter, you can freely adjust the System Prompts under Settings yourself (you can also do this whenever you want, for slightly adjusted results), so just don't mention anything about nsfw and it will be censored, Llama is rather censored by default.

I don't like to use names or anything at all with AI, as immersion is much smoother with a generic name or 'you', but to each their own. 
Cheers!

Created a new prompting template for general prompting settings, roughly four times as fast (Don't quote 😀) to catch onto a desired type of output (counting pages used, not tokens, so 4-5 pages vs 17-20 pages). [Prompter Txt], much more diverse prompts randomly fare better and worse, much more unstable. This works probably rather consistently, only erroring twice with a 7B model @ 8192 context tokens, because of world is as simple of an alteration as possible. 💜

It's a classic AI problem, use a better model, either simply a corporate one, or a finetuned uncensored one. Anything above 12B parameters will run run into this problem less (not impossible, but less likely(just based on my limited testing on various models up to Mistral Small 24B)), but the prompting matters much much more. Not Formamorph's problem, Mistral Nemo is just rather unhinged but good for rp. The default included worlds are... as basic and simple as they get, with no limitations.

World 21 is in a place that's okay for the current size, branching off of the [Slimegirl Mall], there's a few tweaks to... everything😅 that would make it more fun, but 5 nights' work has to end somehow. New format for files, so changes wont override anything super important if someone likes a certain world the way it is 👌(If I remember to save 😝).

[Parasite Choker Mall] with different variations (just whenever I remembered to save 😝), should be simple enough to cut anything out that you don't like if you want to. Flags: [Parasites, Feminization, Slime/Slimegirls(?), Corruption, Pacification, Breast/Thigh/Ass expansion (I really wanted to get the 3D expansion model working for this but that's 3 nights wasted, AI never understood my vision 😿(tiny bit of a engine limit too, documentation for custom models (or everything) would be amazing 🥰))].

All the worlds here: Formamorph Worlds! 

Just found some old (very old, ~2010-2014) .gblorb games, but I've next to no clue how to do them justice in this format. House of Pandemonium (2014 ver) and Gynoid Conversion (~2014?) run through an interpreter like Gargoyle 🪨.

They'd make great worlds, but limitations on Entities, Locations, and System Prompt's leave them largely infeasible to completely reproduce. Edging off of my own set context limits (sub 4k tokens) would make something playable but... not great. An entire re-imagining of the games would be great, but ehhh. Just a thought 😝.

Lol, I was testing 'thea-rp-3b-25r-i1-GGUF/thea-rp-3b-25r.i1-Q4_K_S.gguf', on 120000 max context tokens and .94 temp, and on game start there was a 50/50 chance of weird things happening... But they'd smooth out (after some hidden dialogue) on the next input 😝.

Erm, something happened, maybe? Just randomly.

1. infested_slime_world should be more lenient, but it's probably still a memory limitation. You can work around it by cutting some of the content out of the world (system prompts, entities, etc), or upgrading the AI.

Memory is an inherent problem with Formamorph, the more prompts, Locations, and Entities that are in the worlds, the more that gets sent in an API call every message. Hardcoded, maybe an update could help, maybe not.
You can use higher context models to get around it, I'm recommending at bare minimum 8192 tokens for most of my worlds, erm sorry 😿. Default Formamorph with no changes has 2000 tokens under Max Memory (characters). You should use better (free) models off of OpenRouter, and change the Max Characters for there to be effects 👈.

If you want, I can create custom worlds for anything interesting you'd want, Formamorph is pretty flexible with prompting, Locations, and Entities, it is just the AI that needs to be tailored to your preferences for the best use. I tested with cut-down versions of "Slimegirl Mall" that can run off of smaller memory (just barely usable 😅), those are all the 14XX worlds.❤️

Heyo! Error 400 is typically a client-side error. Does the problem occur in Google Drive or within Formamorph when using the world? If the problem lies with GDrive, could you try downloading the world again? It worked fine with my other alts, and permissions are still the same within GDrive. If Error 400 occurs in Formamorph, the world's probably too much for the AI's context, too many drones to handle.❤️

I mean, it the usual context and AI issue, if you can cope with self-hosting or paying more for higher context or instruct LLMs, they're (almost)always better. The problem with always reinforcing prompts back into the LLM is that it costs more tokens... and can result in bad RP traps thet'll get stuck in.

Made something basic that can be expanded upon with descriptions and other vore types. Relies mostly on how you AI is setup, but somethings can be fixed. Mine always tries to go outside right away... No. 17 Horsey, didn't pick a name oops. 💜

Heyo! Sorta finished my [Slimegirl Mall] [Corruption, Slimegirls, Transformation, Mind Control, (Mostly) Futanari] scenario(s?), minimalist 🏖️sandbox like usual, with vague enough prompting to use worse AI models running smaller contexts. There's a set of (18) singular (mostly futa 💜) slimegirl focused generic scenarios, and a smaller & A regular sized (amount of slimegirls/locations picked) mall scenarios.

This was mostly setup to test using my IPad with a local LLM in Formamorph, so I once again stripped the json images (you could manually delete the images from the json's directly... 😵‍💫). There's also a txt file with almost all the (again, mostly futa 💜) slimegirls I could come up with if you want something different.

Thank you FieryLion ❤️!

Here's the enormous choices I received when I touched the [Choices Prompt]s a bit... I think I need to tweak it a bit more, It's basically serving everything to me on a silver platter, the choices are great but altering them is a hassle with these crazy things.

Just some simple tips to start creating worlds faster;

1. Saving is crucial and you need to always re-download worlds so they save.
2. Try saving your worlds in a proper folder... Lest you want to lose them 😿! I layout the JSON's in one folder, with other folders for world specific images if needed.
3. You can speed up saving by creating simplified names for the JSON files (I use porky_pig_world styling, appending _backup or _V1 for ease of use).
4. You can create multiple versions of worlds for testing easily by clicking the file and appending an additional number to the end of the file name.
5. You can substantially reduce JSON file sizes by swapping out any images you won't see/use with blank PNG files (Easier to use smaller files on tablets).
6. Try running some of your ideas through a stylized AI 👍(or your friends 🤔), they can have really great ideas!
7. Google Drive is a free place to backup and share worlds from, just make sure to change access permissions to [Anyone with the link] and [Commenter] (or viewer... ❤️).

Good Luck! 💜
Worlds (29): GDrive

I'm having a wonderful experience. Thank you FieryLion for this amazing free creative tool, but sorry that it's not being used for expansion fetish content.
It's much easier to create anything you want in compared to anything else I can find (provided that you've either a LLM capable machine or a good API for AI).
I've got to share some of the world's I've setup just for fun (Initially 8, currently 13): GDrive Formamorph
Be careful of prompt size, I very much recommend a minimum of 8192 (more is better) context length (max memory in Formamorph), but sometimes a forgetful AI can be fun to, if you plan for shorter world sessions.

Check out the 'basic template.txt', it is very easy to just get rolling with an idea with a free AI (gpt-4o) and let them use the template to quickly generate super easily tweak-able content.

Thank you, once again, FieryLion, for this great tool.