To go with story generation, AI image generation would be nice for CYOA.
I imagine:
A global property describing the basic style of the story. This can be "fantasy world" but also "children's illustration" or something. This way all the images should look somewhat similar.
Ask GPT to create a prompt from the passage.
Use some(?) API to create images.
Dall-E seems easy to setup
This server can be used to run Stable Diffusion locally. (Can llm.ianbicking.org connect to a local server?)
And then this service to run Stable Diffusion from the cloud.
Midjourney is particularly good at this use case, but has no API.
It's possible that running it in the app will lose too much functionality (or will require recreating too much functionality), but I'd like to at least try.
This requires storing images in IndexedDB and/or JSON base64. Seems like it could get way too big. Not sure if I can just hotlink the images from an API, though that would be better both for hosting and to keep the image with the person who created it.
ianb
commented
1 year ago
To go with story generation, AI image generation would be nice for CYOA.
I imagine:
Midjourney is particularly good at this use case, but has no API.
It's possible that running it in the app will lose too much functionality (or will require recreating too much functionality), but I'd like to at least try.
This requires storing images in IndexedDB and/or JSON base64. Seems like it could get way too big. Not sure if I can just hotlink the images from an API, though that would be better both for hosting and to keep the image with the person who created it.