Writing & Publishing Getting a Book Cover

[Zee]

As I was advised by another author: "Readers do not care if the cover is a scene from the book." So if you want a storm, do a storm. It can be representative.

Click to shrink...

I agree, with reservations—if your characters never go to the beach in the story, for instance, I wouldn’t have a cover with a sea background (maybe that’s because I think too literally about these things?)
And I really dislike cover images of characters who look nothing like the protagonists as described in the story…

But in BK’s case, his characters do end up traveling in the desert, even though there’s no lightning storm, so a desert scene fits the overall tone/atmosphere of the story well, better than the crowded souq scenes, in my opinion, though those were pretty cool.

 

[carolinamtne]

As another literalist, I agree with Zee above.

 

[BKHunter]

I agree, with reservations—if your characters never go to the beach in the story, for instance, I wouldn’t have a cover with a sea background (maybe that’s because I think too literally about these things?)
And I really dislike cover images of characters who look nothing like the protagonists as described in the story…

But in BK’s case, his characters do end up traveling in the desert, even though there’s no lightning storm, so a desert scene fits the overall tone/atmosphere of the story well, better than the crowded souq scenes, in my opinion, though those were pretty cool.

Click to shrink...

Thanks for the comments! I am thinking of adding some heat lightning to some of the scenes. It's not uncommon in the region, although actual rain is, at least in the summer.

<spoiler alert> I'm actually trying to get something close to the scene in the book where the terrorist is watching cloudy skies and waiting for the moon to break.

 

[Zee]

Thanks for the comments! I am thinking of adding some heat lightning to some of the scenes. It's not uncommon in the region, although actual rain is, at least in the summer.

<spoiler alert> I'm actually trying to get something close to the scene in the book where the terrorist is watching cloudy skies and waiting for the moon to break.

Click to shrink...

Alas, poor Khalid!

 

[BKHunter]

As you can see from the last prompt and images I posted, I still have issues consistently getting the man’s arms raised and the moon added. Any suggestions? Using a mask, etc. doesn’t seem to help much. I could always Photoshop the moon in, but the arms would be harder to draw.

 

[Zee]

As you can see from the last prompt and images I posted, I still have issues consistently getting the man’s arms raised and the moon added. Any suggestions? Using a mask, etc. doesn’t seem to help much. I could always Photoshop the moon in, but the arms would be harder to draw.

Click to shrink...

I don’t see any particular reason why he has to have his arms raised…just scanning the sky, looking for the moon would be enough, I think.

 

[M. D. Boncher]

As you can see from the last prompt and images I posted, I still have issues consistently getting the man’s arms raised and the moon added. Any suggestions? Using a mask, etc. doesn’t seem to help much. I could always Photoshop the moon in, but the arms would be harder to draw.

Click to shrink...

You'll have to use a mask I think and keep plugging. Now the below video shows how inpainting and outpainting works with InvokeAI, but the principles still remained the same. It's done at high speed so you can see what's going on in a reasonable amount of time. I was watching it in a start/stop method to see how they were adjusting the prompts. It's very educational. (it also proves how much AI art is not theft.)

 

[Wes B]

It's interesting that you can use previous images as one of the inputs for creating a new one. Is it possible to begin with a sketch of your own, to establish the basic layout and image elements that you want to have? That might give a huge start in getting what you want.

This would seem doubly important for a new, unknown author. I mean, the cover is a BIG determinant as to whether a potential reader will pick the book up or click on the story description. If the author is already known, then the name can be as important as the illustration, and they could afford showing a scene of people merely standing around; the image might only need to establish a mood. For an unknown author, that picture will likely need to convey to the potential reader why this book stands out from the vast sea of "otherness" that's already out there. The more unknown we are, the more heavy-lifting the cover is going to have to do for us.

 

[BKHunter]

Thanks for the responses! Yes, you can use an old image as your "guiding" one and I finally used one of the few that had a moon in it as the new basis, then reduced the amount of freedom the AI had so that most of the new ones created also had moons. Here's one of the latest iterations done this way:



I also got the latest version of Photoshop Elements (my other photo-editing software was obsolete) and will start playing with layers, etc. to get the final image soon. I'm admittedly spending a lot of time doing this, so it may not be cost-efficient, but it's a good learning experience and I'm having fun with it, so there's that.

The masking advice is good, too, and I'll try that again before finalizing it.

 

[BKHunter]

You'll have to use a mask I think and keep plugging. Now the below video shows how inpainting and outpainting works with InvokeAI, but the principles still remained the same. It's done at high speed so you can see what's going on in a reasonable amount of time. I was watching it in a start/stop method to see how they were adjusting the prompts. It's very educational. (it also proves how much AI art is not theft.)

Click to shrink...

Great video; thanks! I wish I could work that quickly.

 

[M. D. Boncher]

Great video; thanks! I wish I could work that quickly.

Click to shrink...

Me too. That's sped up a LOT

 

[BKHunter]

I think this is almost there; it even has the man's arms raised like I wanted. Think of a million monkeys churning out graphics at the same time and you get the idea of how this works. The circular clouds look rather weird; what do you think? I wanted it to be a bit surreal....

:

 

[M. D. Boncher]

I think this is almost there; it even has the man's arms raised like I wanted. Think of a million monkeys churning out graphics at the same time and you get the idea of how this works. The circular clouds look rather weird; what do you think? I wanted it to be a bit surreal....

: View attachment 13830

Click to shrink...

If this is going the way you want, you're going in the right direction. The one thing it needs to do is run through a light img2img pass to give it a more unified look via an artistic style could help. Right now the man looks to be standing out like a piece of bad colormatched clip art. This is a danger of SD images. Sometimes it will do the bad clip art thing without a style or artist to guide it. Fighting with that right now on one of my own images.

The other odd point that isn't working for me is that you have 4 sources of light going on (6 technically with 2 of them being ambient): The moon, the lightning and the "godrays" behind the clouds, plus you have some strong backlighting on the man of a different source and then the golden sunset that seems to be mixing with the sand. All four are making for muddled light. The sand is not reacting to any of these sources. Now much of that can be cured in photoshop fairly simply. They have the training videos out there on how to fix that in layers on youtube. But till those lighting issues get addressed this is an ok cover. Once addressed, it will feel whole and cohesive (at least to my eye).

It's good to make it surreal, but it should still follow the laws of physics to most people's eyes to a big degree lest it subconsciously throw them off.

 

[Zee]

This may be going the way you want, but I’m afraid it doesn’t seem like the right direction to me.

Maybe it’s all the semi-realistic images combined with the crazy colors and sky, or maybe it’s just that your guy looks like somebody’s uncle randomly making dua in the desert…

It’s just like “Wait, what…?”

 

[Wes B]

The circular clouds look rather weird; what do you think? I wanted it to be a bit surreal....

Click to shrink...

The circular clouds, spouting lightning are likely to suggest that some spectacular celestial event is going on, all aimed at the little spot where the lightning ends. If this is not what's happening in the story, it's likely to annoy some readers, unless something equally spectacular is really going on. Surreal can be okay, but be careful that the readers don't end up feeling misled.

 

[M. D. Boncher]

This may be going the way you want, but I’m afraid it doesn’t seem like the right direction to me.

Maybe it’s all the semi-realistic images combined with the crazy colors and sky, or maybe it’s just that your guy looks like somebody’s uncle randomly making dua in the desert…

It’s just like “Wait, what…?”

Click to shrink...

The "Uncanny Valley" is a thing. A very creepy thing.

 

[BKHunter]

I think I was getting too much of an "AI" look by overly restricting DiffusionBee to get what I wanted. I switched to a better program (it was a fairly big hassle; it runs from Terminal) called Automatic1111; here's one of the first images from it, with few restrictions on what it can do other than a suggested guiding image from DiffusionBee. There's a big difference in quality, which is probably partly since I also switched from the 1.5 to 2.1 Stable Diffusion models.

prompt: ((Arab man in ((white robe)) looking at ((crescent moon)) with lightning and birds and storm clouds at night)), ((sand dunes in background)), realistic night sky, by Brooke Shaden and Ted Nasmith and Igor Zonin, trending on artstation, insane level of detail, 8k, cinematic lighting, pastel art

 

[M. D. Boncher]

View attachment 13834

I think I was getting too much of an "AI" look by overly restricting DiffusionBee to get what I wanted. I switched to a better program (it was a fairly big hassle; it runs from Terminal) called Automatic1111; here's one of the first images from it. Big difference in quality.

Click to shrink...

A1111 is right now the closest to an SD "standard" GUI. It can do a ton of stuff that you will be happy for. I just tried training my first ckpt the other day. Mixed results. Still learning. Merging models is fantastic. Their in/out painting is decent. I've done some good pieces with it that I plan to sell soon.

Keep going in the direction you're aiming. you got some nice tools to work with.

 

[carolinamtne]

Why do I have the feeling you have spent as much time coming up with a book cover as you did writing the book?

But this is interesting.

 

[M. D. Boncher]

Why do I have the feeling you have spent as much time coming up with a book cover as you did writing the book?

But this is interesting.

Click to shrink...

NOW you are seeing why I cautioned about the need and time costs. In the end, and this is where artists are going to benefit, you will be better served hiring an artist, even if all they're doing is using AI art in making your cover for you and capitalize on their graphic design skills plus typography work. You're better off focusing on the next book if this is not your passion and able to monetize your skill.

The only reason I've not just chalked this up to "Sunk Cost Fallacy" for me, is that I do intend to have a sideline doing graphic design and this tool will help immensely. Of course, it's been taking months of work to build up skills and understanding and sandblasting the rust off of old skill. But if I'm going to do this, I have to do the work.

(and then being piled on by writing music to go with the novels, plus working with my wife on crafts for crossbranding sales at shows and cons.... Time gets away fast).