Spaces:
Running
on
CPU Upgrade
New release: a few bug fixes
Hello everyone,
a new version of the AI Comic Factory has been released, which includes a few quality-of-life improvements:
- now the latest prompt you typed is saved in the local storage (which means you won't lose your prompt if you reload the page or log-in)
- the single page PDF export should also work correctly
Cool!
- I have a problem. If you have the ai-comic open on two pages, then the prompt specified later is transferred to the older one as well. Without updating. Of course, it generates the same image that I gave previously. But it's a bit annoying.
- The other question. There has already been a program here where the generated images migrated to another person. This can't happen here, only now with Prompt?
- Are there any plans to develop such that you might generate an image based on an uploaded image, that is, in its style or based on its content? If at all possible.
- There has already been a program here where the generated images migrated to another person
It is possible that this type of situation happened in some Spaces using local files as a buffer, or with multiple server instances which got mixed (each space is developed independently by people of various skill levels, so they are not 100% bug proof)
The AI Comic Factory is normally immune to this kind of multiple instance replication issue (images are generated and served on-the-fly and not stored to any disk or database)
- This can't happen here, only now with Prompt?
No don't worry, there is no risk for the prompt: the reason why it is transmitted to your second tab is that because it is stored in the local storage of your browser, which unique for the whole website.
I suppose this could be fixed by allocating a unique ID to each tab, however at the current time the AI Comic Factory works best if there are multiple people using it in parallel, each doing only one request at a time.
The edge case of having one user doing multiple requests in parallel is possible, but can lead to the platform being saturated with requests (eg. an user who might open a new tab to create a comic in parallel, without waiting for the previous one to finish, become frustrated, create even more tabs to go faster etc), reducing the number of people who can freely enjoy AI Comic Factory at no expenses.
- Are there any plans to develop such that you might generate an image based on an uploaded image, that is, in its style or based on its content? If at all possible.
Maybe in the future, the models I tried didn't have a great output quality for comic style (except for generic/bland toon or stock photo styles) but this has improved recently with newer, better models, so I will have to check this year.
It would be nice to have an image for a consistent character or a style indeed, from one image.
Thanks for the reply!