AI Art & Image Generation

2043 readers

1 users here now

A place to share images and art generated by artificial intelligence and similar tools.

Rules:

All posts must be relevant to image generation with artificial intelligence.
Please include the name of the AI or tool used to generate your image at beginning of your post to promote searchability. Example: "[Midjourney] Picture of a lake."
It's not required, but we encourage you to include the prompt used to generate the image in the description of your post.
To avoid spam, please try to limit yourself to five posts a day. Feel free to add as many images to your posts as you'd like.
Please keep NSFW content to a minimum. Risque content is allowed, but pornographic AI art is not. There are plenty of other places to share that. Posts not flagged as NSFW will result in a temporary ban.
Do not self promote your AI tools you created without mod permission. Also any other post about AI tools that seem sketchy will be removed and the user banned at the moderators discretion. Please report a post if you think it should fall in line with breaking this rule.
Please be nice to your fellow users and make sure to follow Lemmy World's rules of conduct: https://mastodon.world/about

Recommended Communities:

Check out [email protected] for more AI Images.
Check out [email protected] for discussion about AI tools you can use in your Dungeons & Dragons games.

founded 1 year ago

MODERATORS

Hyrulian

Conquering the Challenges of AI in Generating Specific Styles and Details (self.ai_art)

submitted 1 year ago by g0zer to c/ai_art

2 comments fedilink hide all child comments

In the realm of AI-generated art, we often find ourselves marveling at the potential for infinite possibilities. However, there's a challenge we frequently encounter: getting our AI assistants to generate specific details or styles accurately... It's like guiding a rebellious brush to effortlessly recreate intricate patterns, where precision battles our creative aspirations.

Have you experienced the frustration of coaxing the AI to produce the exact style or detail you desire? Whether it's a particular brushstroke, a specific lighting effect, or even a distinct color palette, let's discuss the roadblocks we face and discover strategies to overcome them.

As we delve into the depths of AI-generated image generation, what aspects have you found most difficult to tackle? Share your experiences and insights with the community. Together, we can unravel the complexities and develop techniques to harness the full potential of AI in creating the art we envision.

Here are a few tips to kickstart the conversation:

Fine-tuning Parameters: Experiment with adjusting the parameters and settings of the AI model. By tweaking the input variables and exploring the available options, you may discover the sweet spot that unlocks the desired styles or details. Weighting & grouping of terms plays a huge part here, IMO.
Iterative Feedback: Engage in an iterative process with the AI model. Provide feedback and guide it towards the desired outcome. By iteratively refining the generated results, you can learn the intricacies of the AI model to better understand and replicate the specific styles and details you seek. I often run a prompt 20 or 30 different ways before I hit the style I'm looking for; taking feedback from the model in the form of my image generations.
Artistic Fusion: Take inspiration from multiple artists or artistic movements and experiment with merging their distinct styles. Fuse elements from contrasting genres, periods, or techniques to create something entirely new and mesmerizing. It's in these unexpected juxtapositions that remarkable sparks of innovation often ignite. For example, reimagine a classic impressionist landscape through the lens of surrealism or infuse elements of street art into a traditional portrait.

I know a specific style that has given me a very hard time is creating different subjects in the style of old Roman marble statues. I have this list of people and events I want to see carved into marble. Can't get the model to reliably pump out results though, it always ends up getting surreal or "uncanny valley" quickly.

What tips or suggestions can you give the community? What problems are you running into? What styles can you just not get to work?

top 2 comments

sorted by: hot top controversial new old

[–] Hyrulian 2 points 1 year ago (1 children)

Thanks for sharing your tips with the community! I really should have read the Docs before fully jumping into Midjourney because there are so many arguments you can add to drastically change your generations that I never would have known if I didn't look!

[–] g0zer 1 points 1 year ago

There really is a wealth of info just on the Midjourney site and they often update the documentation as new models roll out, so it's a good idea to revisit them from time to time.