Colossus Krea Workflow

Unlock a new level of control with this workflow, meticulously made for the Colossus Krea Merged checkpoint. It's not just about generating images; it's about transforming your ideas into stunning, high-quality visuals.

Through my own experimentation, I've found that using a dual CLIP loader is the most effective way to manage both positive and negative prompts. This setup allows for a nuanced approach to guiding the image generation process. The clip_l text encoder is best used for standard prompting, describing the subject, style, and general composition of your desired image. Meanwhile, t5xxl excels at capturing complex, long-form descriptions and is ideal for articulating highly detailed or intricate concepts.

Here's how to apply this technique:

• CLIP_L: Use a comma-separated list of keywords. Think of these as the main tags for your image. This is where you would place things like (portrait photography, studio lighting, highly detailed skin texture, vibrant colors, bokeh).

• T5XXL: Use a full, descriptive sentence or paragraph. This is where you describe the scene, the feeling, and the narrative. For example: A stoic, determined young adventurer stands atop a moss-covered cliff overlooking a majestic, sun-drenched valley, with a single ray of light illuminating their face.

By separating these two types of prompts, you can achieve a level of control and detail that is otherwise impossible. For your convenience, the negative prompt is already configured with an extensive list of terms to automatically filter out common unwanted artifacts, so you can focus on what you want to create without having to worry about what you don't.