I’ve got another request if it’s not too much trouble, one of only Maugrim.

Specifically, this description of him from chapter 12: “a wolf standing on its hind legs, with its front paws against the tree-trunk, snapping and snarling. All the hair on its back stood up on end.”

Again, if it takes up too much of your time, don’t worry about it. Just thought I’d ask.

@aslanthelion : In the book, it describes Susan as being close in the tree above Maugrim.

Then he realized it was a wolf - a wolf standing on its hind legs, with its front paws against the tree-trunk, snapping and snarling. All the hair on its back stood up on end. Susan had not been able to get higher than the second big branch. One of her legs hung down so that her foot was only an inch or two above the snapping teeth.

Sorry for the dumb question, Do you want:

a close up of the wolf (so that you can not see Susan in the second tree branch above the wolf),

a Susan’s POV shot of the wolf below her snapping up at her, 

a shot of Maugrim with Susan’s leg in the shot, implying the danger,

ideally both Maugrim and Susan (but trying to keep it simple to save time and make it easier for AI and hence asking for only Maugrim - thanks by the way, if this is the case),

some combination of the above,

something else that I haven’t thought of?

I’ll see what I can do 🙂

@aslanthelion 
Close up of Maugirm:  MaugrimFromSide.jpg

Susan's Point of View:  MaugrimBeneathTreeB.jpg

Maugrim attacking Susan:  SusanAndMaugrim.jpg

Maugrim attacking Susan:  SusanAndWolf.jpg

Maugrim attacking Susan:  MaurgimAttacksSusan.jpg

Hey, sorry for the late response.

Wow, thank you for all of these! I was hoping for one of just Maugrim without Susan, but I appreciate you going to the trouble of trying different perspectives!

Unfortunately none of them are exactly like I meant...I was specifically hoping of one from the side without Susan, but that shows all of Maugrim, including his feet. I'm sorry I'm so picky! But still, thank you so much for them all! I'd have to say the first two are my favorites since they have just Maugrim, but they're all good. His design in the last one is really cool, he looks bulkier and more like I imagine him to.

@aslanthelion 
More side shots of Maugrim:

Posted by: @davidd

@aslanthelion 
More side shots of Maugrim:

 MaugrimFullA.jpg

 MaugrimFullB.jpg

 MaugrimFullC.jpg

 MaugrimFullD.jpg

 MaugrimFullF.jpg

Oh wow, thank you!

By the way, what program are you using for this? I might want to try it myself!

Hi @aslanthelion 

I use a few different applications.  I am a cheapskate, so I use the free versions, however pretty much everything below has a paid version if you want more functionality.  If you do want to pay, I would recommend trying each one in its free form first and once you have a good handle on what works for you, move to a paid version:

1. Chat GPT: I use a free account (limits you to 4 or 5 image generations a day, but it is very good at following your prompt)
https://openai.com/chatgpt/overview/.   You generate from the website.  It can do both text-to-image and image-to-image (I.e. adjust an existing image). It has very good prompt adherence.

2. Microsoft Co-pilot.  If you have a windows computer, you can use Microsoft Edge.  Edge will have a built in ChatBot, you can send it requests to paint pictures in a similar manner to ChatGPT.  (I think this is an older version of ChatGPT) still has a limit of generations per day on  a free account.  Alternatively, you can use bing ( https://www.bing.com/) and bing has the Microsoft Co-pilot chat.  You generate from the chat bot on the website.  Again, you get a limited amount of generations with a free account, but I've managed with a free account.

3. Google imagen.  This is my favorite image generator to use.  ( https://labs.google/fx).   You generate from the website.  You need to register an account, but you can generate 6 images per minute (which is more than enough) on a free account.  It only supports text-to-image, but it does very good text-to-image.  This is a good one to experiment with as you have almost an infinite number of image generations to find what you want.

4. Deep Dream Generator.  ( https://deepdreamgenerator.com/) Generate from the website. The prompt adherence of this one is not as good as the rest, however it excels at generating more artistic images.  If you want to experiment with abstract, renaissance, or other popular styles, this one can produce results that exceeds the others. The free version allows for several generations per day (it used to require you to get about a 1000 likes for you to be added in the free category where you have 70 credits per day, but I THINK they now give that to you as soon as you create an account).

5. The experimental version of Google Gemini's 2.0 with images can generate images.  It can also edit existing images.  The in-painting of this tool is nowhere near as good as Chat GPT, but you can do unlimited edits to images (as opposed to ChatGPT, which only allows 4-5 generations.

6. Often these tools do not produce exactly the result that I want (either in terms of a specific detail) or the style.  I use ComfyUI with the Stable Diffusion XL model.  ComfyUI supports more advanced models than SDXL, but SDXL has (in my opinion) the best ControlNet integration.  Control Net gives a lot of control over ensuring that details of the original image that you care about can be maintained in the updated version.  I.e. Pose Control nets analyze the position of arms, legs etc in an original image and ensure that a person is positioned in the exact same position with the same pose (though you could change the person from a Caucasian man to an Asian woman and the generated Asian woman would be in the identical position in the same pose, with the same expression on her face).  It has other controls to ensure the outline is maintained, etc.  It's very powerful.  ComfyUI and Stable Diffusion XL also support a lot of Loras / art styles, so it gives a lot of flexibility to customize your images in terms of style.  (I have adjust a large number of the images that I've generated to give it a finish that is between watercolor and oil on canvas because I like that look.)

7. Finally I use some sort of image editing tool (I use Paint.Net because I can use it for free; though I'd use photoshop if I paid for the license as photoshop is very good for compositing and color grading - but pretty much any image editor can do the job.  Even MS Paint can do compositing now that they support layers).

Hope that is helpful! Have fun 🙂

Posted by: @davidd

Hi @aslanthelion 

I use a few different applications.  I am a cheapskate, so I use the free versions, however pretty much everything below has a paid version if you want more functionality.  If you do want to pay, I would recommend trying each one in its free form first and once you have a good handle on what works for you, move to a paid version:

1. Chat GPT: I use a free account (limits you to 4 or 5 image generations a day, but it is very good at following your prompt)
https://openai.com/chatgpt/overview/.   You generate from the website.  It can do both text-to-image and image-to-image (I.e. adjust an existing image). It has very good prompt adherence.

2. Microsoft Co-pilot.  If you have a windows computer, you can use Microsoft Edge.  Edge will have a built in ChatBot, you can send it requests to paint pictures in a similar manner to ChatGPT.  (I think this is an older version of ChatGPT) still has a limit of generations per day on  a free account.  Alternatively, you can use bing ( https://www.bing.com/) and bing has the Microsoft Co-pilot chat.  You generate from the chat bot on the website.  Again, you get a limited amount of generations with a free account, but I've managed with a free account.

3. Google imagen.  This is my favorite image generator to use.  ( https://labs.google/fx).   You generate from the website.  You need to register an account, but you can generate 6 images per minute (which is more than enough) on a free account.  It only supports text-to-image, but it does very good text-to-image.  This is a good one to experiment with as you have almost an infinite number of image generations to find what you want.

4. Deep Dream Generator.  ( https://deepdreamgenerator.com/) Generate from the website. The prompt adherence of this one is not as good as the rest, however it excels at generating more artistic images.  If you want to experiment with abstract, renaissance, or other popular styles, this one can produce results that exceeds the others. The free version allows for several generations per day (it used to require you to get about a 1000 likes for you to be added in the free category where you have 70 credits per day, but I THINK they now give that to you as soon as you create an account).

5. The experimental version of Google Gemini's 2.0 with images can generate images.  It can also edit existing images.  The in-painting of this tool is nowhere near as good as Chat GPT, but you can do unlimited edits to images (as opposed to ChatGPT, which only allows 4-5 generations.

6. Often these tools do not produce exactly the result that I want (either in terms of a specific detail) or the style.  I use ComfyUI with the Stable Diffusion XL model.  ComfyUI supports more advanced models than SDXL, but SDXL has (in my opinion) the best ControlNet integration.  Control Net gives a lot of control over ensuring that details of the original image that you care about can be maintained in the updated version.  I.e. Pose Control nets analyze the position of arms, legs etc in an original image and ensure that a person is positioned in the exact same position with the same pose (though you could change the person from a Caucasian man to an Asian woman and the generated Asian woman would be in the identical position in the same pose, with the same expression on her face).  It has other controls to ensure the outline is maintained, etc.  It's very powerful.  ComfyUI and Stable Diffusion XL also support a lot of Loras / art styles, so it gives a lot of flexibility to customize your images in terms of style.  (I have adjust a large number of the images that I've generated to give it a finish that is between watercolor and oil on canvas because I like that look.)

7. Finally I use some sort of image editing tool (I use Paint.Net because I can use it for free; though I'd use photoshop if I paid for the license as photoshop is very good for compositing and color grading - but pretty much any image editor can do the job.  Even MS Paint can do compositing now that they support layers).

Hope that is helpful! Have fun 🙂

Awesome! Well, thanks for sharing!

I tried to make some of the same shots of Maugrim, they're still not as good as some of yours, though.

In ChatGPT, one of the prompts I tried was:

"Create image painterly fantasy illustration in the style of traditional watercolor and gouache of a menacing, large, grey, stocky wolf with sharp teeth bared (saliva between the teeth) and front paws against the trunk of a tree, reaching up the trunk, looking upwards ferociously. All the hair on the wolf's back and tail are standing up. The wolf's hind legs stand on the ground. The forest consists of tall, slender trees with gray-brown bark and sparse green foliage at the top, set against a muted green and gray background suggesting a dense woodland.
**Art Style:**
Medium/Technique: a traditional media, likely watercolor or gouache, due to the soft blending, atmospheric quality, and visible brushstroke textures, particularly in the sky and landscape.
Rendering: The rendering should be detailed but stylized, falling under illustrative realism rather than strict photorealism. The overall feel should be painterly, but characters are clearly depicted with expression and costume detail.
Atmosphere: Should have dramatic lighting and the soft blending of colors.
Painterly Quality: Should have a hand-painted look via visible textures and blended areas.
In summary, it should be a painterly fantasy illustration, using traditional painting techniques like watercolor or gouache.
Prefer use of warm colors"

I think I then edited the final image in paint.net, but that should give you some idea of the kind of prompts that ChatGPT and imagen respond well to.

If you find an image that you like - it often works well to ask ChatGPT or Gemini to describe the image in terms of style / content and they can give you a lot of the details.  You can then incorporate those details into your prompt.  Give it time, it takes some experimenting to find what works for you  

@davidd 

That's great, thank you!