Thank you for feedback, Osanaiko. Some of your points resonate with my vision for the product .
In short :
In my mind, the product should be something like an infinite story machine, where users can input background information and characters, then press 'start' and AI starts generating a few sentences of the story. After that users can select AI-generated options to continue the story (like dialogue options in visual novel) or insert their own text to continue. Additionally, users can press the 'visualize' button to get an AI-generated image based on the selected text/part of the story.
Sharing stories, the ability to read other users' stories, etc., can be implemented as needed
It can be done using multiple small AI models or a few larger ones, nothing is impossible in this regard. Main issue is the hardware costs: big models require 40+ GB memory graphics cards, small runs on consumer grade cards but they must run simultaneously because small models are narrowly focused, so it costs less than the big one, but still more than $500 per month due to multiple instances. I'm not talking about any other expenses like AI training or software development, only AI inference costs.
I have some savings to sustain development and hosting for a couple of months (i believe it would be enough to release MVP) but not sure if anybody is interested in that kind of product.
Looking forward to hearing what yo all think, pals!
P.S. glad that u enjoyed the chatbot!
In short :
In my mind, the product should be something like an infinite story machine, where users can input background information and characters, then press 'start' and AI starts generating a few sentences of the story. After that users can select AI-generated options to continue the story (like dialogue options in visual novel) or insert their own text to continue. Additionally, users can press the 'visualize' button to get an AI-generated image based on the selected text/part of the story.
Sharing stories, the ability to read other users' stories, etc., can be implemented as needed
It can be done using multiple small AI models or a few larger ones, nothing is impossible in this regard. Main issue is the hardware costs: big models require 40+ GB memory graphics cards, small runs on consumer grade cards but they must run simultaneously because small models are narrowly focused, so it costs less than the big one, but still more than $500 per month due to multiple instances. I'm not talking about any other expenses like AI training or software development, only AI inference costs.
I have some savings to sustain development and hosting for a couple of months (i believe it would be enough to release MVP) but not sure if anybody is interested in that kind of product.
Looking forward to hearing what yo all think, pals!
P.S. glad that u enjoyed the chatbot!
Last edited: