If you've read one of my earlier blog entries, you know that I'm not a fan of current AI. I believe that it has another 30 years to become truly useful. Right now, it's like a 7 year old that will usually do what it's told but could go wild at any moment.
That said, I'm learning how to use Google Gemini to turn some stories into some sort of visual storytelling, whether it's illustration, like a graphic novel/comic book or video.
What I've learned in the last few days is nothing is permanent but the frustration.
Characters do not remain consistent. Gemini does not have a way to keep a character pinned to your story. There is no project, either, although they seem to be making attempts at projects.
I would use a high school or university aged character and in a subsequent image, they've aged 20 extra years. It's like writing a story around the SIMS.
I've seen some people's work on YouTube and I know that they're using better software that has better control. Someone was discussing with me why I chose Gemini and I didn't have a good reason, but that it was available to me. He mentioned two other products and I need to find them and see what they do.
I've spent a lot of time with the camera, capturing a moment in order to tell a story. Is it possible to switch platforms and media to something purely data-driven?
I shall continue to try. Wish me luck.
Update 2026.06.03: There are these programming or scripting additions called Gems. There is also Memory, which I assume is a database of references. It says "Import into Memory", which sounds like you have other tools to help you create that references.
Google mentions a Storybook Gem that allows you to build 10 page illustrated books with includes languages translation. I'm not sure whether it's speech, text, or both. Ten pages will be my first chapter, if I can squeeze it into smaller frames, maybe.
As a former software developer, maybe I can learn to make Gems. I know. I should be able to build my own software to do whatever I want. It doesn't always work that way.
Update 2026.06.05: Little success, server problems, too much activity, hidden problem, little success. Finding every malfunction possible!
I was a good software tester as a software developer. I have found so many snags in Gemini that I cannot count them all. One thing, I had a discussion about my story with the conversation specialist.
We "talked" about a lot of ideas to implement and it's a very quick brainstorming tool. Whether all that got put into a "memory" or not, I don't know. For all I know, being Google and all, they'll sell my idea to someone else. I have to start somewhere.
For image generation, someone mentioned Stable Diffusion. They must have trained it on magazines they got at garage sales. It had no idea in many cases and only created one leg or one arm for certain situations with people. It has potential, but I know that Google and other companies training AI burned through billions of dollars, billions of books, newspapers, magazines, music albums, TV shows, movies and more.
Maybe, the conversation specialist will actually work.
Update 2026.06.08: I downloaded Claude by Anthropic. It is more work-related and on submitting the text of my story, it was able to make some valuable insights. Perhaps, I can use a combination of Gemini and Claude\to get there. There was a big difference between Claude and Gemini on analyzing the story. Claude created a .docx document with my text and its observations, included and placeholders for expansion.
I also submitted the same text to Gemini and it was good. The character generation must arrive last, for the sake of consistency. At this point, I'm not even sure that the characters would last through a text revision.
I tried the Copilot app on my Android phone and the image generator was willing to generate human-like images. The protection layer was not. I removed the Copilot app. I tried Copilot on my new Windows machine and it worked similarly. It did not generate anything, even when I used the prompts it gave me.
Update 2026.06.20: I've been using for a while now. I thought that it was 5 weeks since the first time I tried it and the last 2.5 weeks have been intense.
Gemini can't tell between my story-weaving in the scenarios and has given me a listing of numbers to call for help. Although it's an annoyance, it's a badge of honor concerning my abilities. I'll take it, even though it drivers me crazy.
Gemini is a little too helpful sometimes and not enough other times. It uses Google Maps reliably and shows me where my "virtual" trip is going. I can appreciate that, especially if I ever go to one or two of these places.
I've seen a lot of the USA, even by the time I was three years old. The U.S. Navy does that to a family.
When I'm going head-to-head with Gemini in a historical flashback, I think Gemini is surprised by me. I'm not surprised that it's been fed so much information. I'm surprised that I remember so much.
Now, the problem is that there was a scenario I stopped last week and something happened and it's trashed. I couldn't recover it and it doesn't automatically read through and upgrade it. It wasn't a big deal but it was somewhat important. I'll survive.
I've had some other strange experiences where the scenario will suddenly loop, especially if there is image generation because the generator component can forget to clear data on occasion. It's good on short scenarios or just creating images and leaving. I'm currently running a tourism advertising scenario and it's been good until the image generation at the final site. I don't know the place personally, so I needed some assistance to get me to the images I wanted to get. It changed my model, moved things all over the place. Continually gave me different characters, and freaked out here and there. So did I.
It took a lot to get through 6 days and to see it have a minor meltdown signaled the worst. I did not want to start again. I still don't. I'm hoping that some incremental work outside of the scenario and then, adding it carefully will make things work. It's almost finished.
As I've mentioned, the 7 year old child is alive and acting out, as expected.
No comments:
Post a Comment