[AI Image Generation] The fun and the rabbit hole of image generation discovered by thoroughly mastering Stable Diffusion.

Nice to meet you, or hello!

I'm Ikusuma.

Following my previous self-introduction article, this time I'd like to write about how I've been thoroughly mastering "Stable Diffusion (AI image generation)," which I've been obsessed with every day.

While subscribing to web-based illustration generation services is fine, I prefer a local environment and use it every day.

For those who are wondering, "How much can you really obsess over AI image generation?", I hope this conveys at least a little of the charm of that deep rabbit hole!

Why "Stable Diffusion in a local environment"?

Cloud generation services are convenient, but the biggest appeal of building and mastering your own local environment lies in the "overwhelming freedom and control."

• Generate as much as you want without worrying about limits or quotas

• Freely combine and test the latest models, LoRAs, ControlNet, etc.

• Pursue generation speed, resolution, and prompt fine-tuning to the limit

The process of running your PC's graphics card (GPU) and memory at full capacity to chase that ideal image has a certain masculine romance, much like customizing and tuning a self-built PC.

How I "thoroughly master" it with my usual obsessions

I don't just type in prompts to output images; I conduct experiments and verification every day on points like the following.

• Verification of combinations of models (Checkpoints) and LoRAs

  • Comparing how much I can control the touch, art style, and texture expression while keeping logs.

• Complete control of composition using ControlNet, etc.

  • The work of crafting an image exactly as intended, rather than leaving poses and composition to chance.

• Pursuit of upscaling and details

  • Exploring processes to prevent degradation when increasing resolution and to raise the quality of even the finest textures.

Rather than the time spent looking at the generated images, the process of verifying "how I can get the quality I want even more" has become a hobby in itself (lol).

The "true fun of image generation" I understood after using it

What I felt after thoroughly mastering Stable Diffusion is that "AI is not just an automatic generation tool, but a powerful partner that expands my own expressive power."

The more you have your own specific vision, such as "I want to create this kind of worldview" or "I want to express this texture," the more effective your ingenuity in settings and combinations becomes.

Of course, the sense of a breakthrough when an unintended, miraculous image suddenly appears is the real thrill of running it locally.

Things I want to write about on note in the future

From now on, I plan to share some of my "Stable Diffusion mastery know-how" on this account, such as the following!

• Personal experiences with local environment setup and performance optimization

• Verification logs of "this setting or method worked well"

• Behind-the-scenes stories of creating works using generative AI

I hope this will be helpful for those who "want to try Stable Diffusion locally" or "want to enjoy AI image generation more deeply."

If you found this article helpful, please be sure to "like" and follow me!