Looking for feedback: An AI tool to automate in-chapter illustrations

[ThinkOfSomeCoolName]

This is my first post here, so hello!

You know how sometimes authors include art of key scenes within their chapters? I thought it’d be neat to automate that process, so here we are! I’ve been working on a free tool where you can simply paste your entire chapter, and the site will identify key moments and generate an AI prompt for an illustration using the standard guidelines.

Then, you’d be able to generate beautiful illustrations of your favourite moments. The whole process would hardly take a minute or two.

Preview: Illustrate AI

I’d love to hear your feedback or any advice you may have!

 

[Comiak]

I gave it a try, and I like the concept, but it is a little lacking in some places. A 3000-character limit is entirely too short for a chapter, but I can understand that the AI may get overwhelmed if there's too much. I think this is more a tech limit so I'm not going to nitpick too much on the character limit. However, rather than just failing to analyze the chapter, it should specifically say that the limit has been exceeded because, at first, I thought the limit was word count, not character count, so I went over the limit and was confused as to why it wasn't working.

The feature that picks out key moments is surprisingly good and is something a lot of AI adventure/story sites are missing. Those sites have a sort of memory system, but that memory system cannot identify plot points as well as what you have. The only complaint I have is that I could be refined more, but I'd rather it flag unimportant moments than miss good ones.

The major complaint I have is the art generator. From my own experience, all AI art generators that turn a story into a picture are complete garbage, and that's being generous. Turning prompts into art is one thing that can be done well, but AI lacks conceptual knowledge of things. For example the picture I wanted was a person pulling an amulet from a crumbling statue. The AI fails to depict a statue that is falling apart because it doesn't actually know what a statue is and how it could be falling apart; all it knows is that if someone asks for a statue, it should show picture based on its training. In other words, the AI is good at repeating what it's taught but can't apply its 'knowledge' in new and abstract ways which is what the AI needs to be better at this specific task.

I am sure I could have gotten something close-ish to what I wanted, but as far as I can tell, there is no way to reroll the generated art, and trying to generate it again only resulted in an error. This is another major problem because AI just isn't that good and is really just luck on whether or not you will get something good. So having to refresh the page and paste the prompt back in is just a really bad way of doing things.

My advice is probably to double down on the tech that finds key moments, as that is something everyone else is failing miserably at. Consider selling it or licensing it to other story or adventure AI sites unless you can make a better art generator, as that is the main problem you have.

 

[ConansWitchBaby]

I of course. Immediately went for smut. It gives me nothing.

 

[SoaringMoon]

This is my first post here, so hello!

You know how sometimes authors include art of key scenes within their chapters? I thought it’d be neat to automate that process, so here we are! I’ve been working on a free tool where you can simply paste your entire chapter, and the site will identify key moments and generate an AI prompt for an illustration using the standard guidelines.

Then, you’d be able to generate beautiful illustrations of your favourite moments. The whole process would hardly take a minute or two.

Preview: Illustrate AI

I’d love to hear your feedback or any advice you may have!

I know someone programming this exact tool at an enterprise scale.

 

[TreasureHouse]

I of course. Immediately went for smut. It gives me nothing.

I got error messages, but I tried feeding it a scene where ghoul sprouts 10 dongs and violates an entire sleeping campsite.

I don't know where I went wrong with it

 

[CharlesEBrown]

Came very close to what I had in mind on the fifth try, then crashed every time I tried to refine it after that (first try was hilariously off due to leaving too much detail).

 

[RepresentingCaution]

It's not generating images for me, but I like the overall concept.
What I would like even better is to be able to train the AI to generate images of the characters one at a time before getting to the group images with multiple characters so there will be less to refine in the final stages.

 

[CharlesEBrown]

I got error messages, but I tried feeding it a scene where ghoul sprouts 10 dongs and violates an entire sleeping campsite.

You confused it - you told it to do what AI art generators usually do anyway and it just did not understand the request.

 

[ThinkOfSomeCoolName]

Thanks of your reviews! Just to clarify:

It's not generating images for me, but I like the overall concept.

1. The art generator is a paid service, so sadly until I have a clear solution for the cost, it’s going to be limited.

A 3000-character limit is entirely too short for a chapter, but I can understand that the AI may get overwhelmed if there's too much.

2. 3000 is a word limit not a character limit. It should be sufficient for most chapters.

I know someone programming this exact tool at an enterprise scale.

3. Damn, so much for that... But well, I haven't tried the site yet (It needs an invite code), but mine's more of a key moment extractor + prompt crafter + Image generator, so there must be some differences.

My advice is probably to double down on the tech that finds key moments

4. Noted


Please leave more reviews and feedbacks!

 

[WinterTimeCrime]

It's been analyzing over the course of two league matches... I'm about to crash out ??

 

[ThinkOfSomeCoolName]

It's been analyzing over the course of two league matches... I'm about to crash out ??

Sorry, but can you please explain where you are experiencing this issue? Message me personally if that's convenient. Coz it's working smoothly for me...

 

[WinterTimeCrime]

Sorry, but can you please explain where you are experiencing this issue? Message me personally if that's convenient. Coz it's working smoothly for me...

I apologize. I was being stupid. Web-based ML algo was refreshing my tabs.

This is pretty cool. I can definitely see cinematographers using this tool for story boards and stuff. For someone who has a more ranobe style of writing, this would be very useful for storytelling (especially for us wanting to break into the LN category, though far-fetched).

The prompts did tailor some decent pictures, albeit unusable (but it's not the prompt's fault; it's AI-gen. What are you gonna do?)

Great joooob! You get a gold sta- cookie!

 

[Jocelyn_Uasal]

This is my first post here, so hello!

You know how sometimes authors include art of key scenes within their chapters? I thought it’d be neat to automate that process, so here we are! I’ve been working on a free tool where you can simply paste your entire chapter, and the site will identify key moments and generate an AI prompt for an illustration using the standard guidelines.

Then, you’d be able to generate beautiful illustrations of your favourite moments. The whole process would hardly take a minute or two.

Preview: Illustrate AI

I’d love to hear your feedback or any advice you may have!

Oh wow I've actually never considered having chapter illustrations until now, I really like that idea! I'd use it either as a placeholder or as reference pieces to give to an artist.

-
I went ahead and tried it out, it's actually pretty competent, way more so than I expected. It says there is a 3k character limit but I pasted my entire chapter (about 20k characters) and it didn't seem to care.

I took a part from my first chapter to make the image below, I'm pretty happy with it other than the 6 fingers lmfao

 

[LoneQuack]

Honestly, while not what I had pictured in my mind, and a tad bit simplistic and generic, it was actually pretty cool. Won't be using the current art but will probably play with it for a while.

 

[CharlesEBrown]

Honestly, while not what I had pictured in my mind, and a tad bit simplistic and generic, it was actually pretty cool. Won't be using the current art but will probably play with it for a while.

Yeah, if I could get my graphics program installed I would very likely have kept two of the five images it generated (one was utter rubbish and two just not as good) to use as reference images, but without that installed, what it came up with was, for now, useless to me. Still closer than I had expected.

 

[SoaringMoon]

mythiq doesn't do key moment extraction, but does do image generation, prompting, automatic writing, text to speech, and also give unique voices to each character

I can dm you an invite code.

 

[CharlesEBrown]

After several near misses (and even more ... well, maybe a block away from the ballpark) this came up with a pair of images I might be able to use one of (or bits and pieces from)...

Still need some tweaking but might be able to do something with these.