Maybe start off by using the right terms and the right context in the description.
I doubt that anyone would decline their lifesaving medicine that were developed with AI. Or all the decades old software tools for the disabled.
AI is a very broad and old field in computer science, the more recent aspects of LLM and image generation, specifically in digital RPG products, are often viewed negatively. And with good reason. Just as many of the self made products have a negative perception, that goes even back to the D20 days where everyone and their father were publishing D20 products in often abysmal quality...
There are multiple reasons why I buy stuff, besides a collectors drift that you might call an addiction...
- Need: Like food, shelter, etc.
- Pleasure: Good food, something nice to look at, a good book, etc.
- Entertainment: Something that isn't necessarily at the high level of 'pleasure', but something to enjoy nonetheless. A (RPG) book, a movie, a series, a copmputer game, a board game, etc.
- Time-savers: Better computers, specific software, tools, appliances, services, etc.
LLM and image generation falls under the header of time-savers in my book, which as an end-function have an entertainment objective. Example:
I've been using LLM (DeepSeek) to generate room descriptions for a mega dungeon, I could write those myself, but it saves enormous amounts of time to use LLM, and to be honest LLM might be better then 90% of the people writing room descriptions for adventures (heck a LOT of what I got out of the LLM is better then the official TSR/WotC descriptions). But what people forget is that it still takes time, if you're doing a mega dungeon it still takes a LOT of time.
From that LLM room description I use a deep learning text-to-speech service (also AI => Elevenlabs) to convert it to audio files. That requires a subscription, which isn't cheap in the first place.
I also use that room description with another LLM (ChatGPT with a specific instruction set) to generate image prompts for Midjourney image generation, so with each room I have a text description, an audiofile so I don't have to read it aloud myself, and an image to show the players.
Over the last year and a half I've spent almost $500 on just text to speech and image generation services. The LLM stuff I used was free (but can also be quite expensive) and I don't know how many hours on finagling prompts, setting up the right settings, input for text to speech, organizing, etc. BUT if I had done that completely myself (writing and audio recording) would have probably taken a decade if not decades. Plus more money in audio recording hardware/software, audio dampening panels, etc.
Another tool I bought is MacWhisper, which can use a local LLM model (Whisper Large V3 Turbo) to convert speech to text. Everything is done locally, so nothing goes into the cloud. I record our RPG sessions (everyone is aware this happens) and use the software to transcribe the audio tracks and it creates a text file with who says what that it somewhat accurate. Doing that by hand would take a LOT of time, at least as long as the 4 hour recording, probably a LOT longer... I could also pay someone to do that, but that would be very expensive. Now the cost is a single $70 payment and a strong enough local machine (Mac Mini M4 Pro 64GB, but would work with less).
Would I have payed someone for a good finished AI made product that had done all that, heck yes! Not because I could not do it myself, but because I would assume it would be cheaper then the subscriptions I paid for and the oodles of time I spend on it...
People who say "I'd generate that myself!" they clearly never have.
That doesn't mean that I would accept any completely LLM/IG generated piece of $#!& without looking at it. But if it does what I want at a reasonable price, I'm fine with it. After all I've been buying TSR/WotC products for decades...
Sidenote: DTRPG allows for pdf previews, when they are not present without some darned good reason (like the product isn't a pdf), then that's already a strike against the product, no matter if it's made via LLM and image generation or not. If there are no reviews, no previews, then it's a pass no matter how it's made.
