How Fine-Grained Control Works in ProsePainter 2

Fine-grained control in ProsePainter 2 comes from three layers working together: text prompts that describe what the AI should add, an AI strength setting that decides how much the AI overrides your own brushwork, and ControlNet guides (edge and depth maps) that constrain the shape and structure of what gets generated. The practical goal is artist-directed output — you paint, and the AI fills in around your decisions rather than replacing them.

The three control layers

Text prompts: direction, not command

ProsePainter 2 is framed as "Paint — don't prompt," but text prompts still exist as a steering tool. According to the site, you use prompts "to guide the AI in your generations." The distinction matters: a prompt here is one input among several, layered on top of what you've already painted, rather than the sole source of the image.

Prompts can also be attached to style ingredients, letting you add "custom prompts for more control" when building a brush. That means the same prompt logic applies at two levels — the overall generation and the individual style component.

AI strength: the balance dial

The AI strength setting is the core of fine-grained control. The site describes it as adjusting "the balance between your input and AI creativity."

• Lower strength keeps more of your original strokes and lets the AI make smaller, more conservative additions.
• Higher strength gives the AI more room to reinterpret and invent.

This is the setting to reach for first when results feel either too timid or too disruptive. If the AI is smothering your composition, reduce strength; if it's barely responding to your prompt, raise it.

ControlNet guides: structural constraints

Edge and depth maps are the most precise lever. The site lists "advanced guides like edge and depth maps to steer your generations with precision."

• Edge maps lock in outlines and contours — useful when you want the AI to respect the boundaries of a drawn shape.
• Depth maps encode near/far relationships — useful when you want generated detail to follow the spatial logic of your scene.

These guides constrain where and how the AI applies its creativity, which is different from the strength dial, which controls how much.

How the layers combine

A useful way to think about it: prompts say what, strength says how much, and guides say where.

For example, if you've sketched a figure and want the AI to render fabric texture without redrawing the silhouette, you'd set an edge map from your sketch, write a prompt describing the fabric, and keep AI strength moderate so the texture fills in without the model reinterpreting the pose.

Style controls as a fourth input

Style ingredients and image-reference brushes also feed into the generation, and they interact with the controls above. The site notes you can "mix several images to create hybrid styles" or "capture a section of your work-in-progress to lock in a style." A style built from your own work-in-progress is effectively another way of keeping the AI consistent with your existing direction — complementary to a lower AI strength setting.

Practical starting sequence

1. Paint the composition you want to keep.
2. Add a text prompt describing the detail or content the AI should contribute.
3. Set AI strength conservatively, then increase only if the AI under-delivers.
4. If structure drifts, add an edge or depth guide rather than pushing strength higher.
5. Refine style ingredients last, since they shift the overall look across the whole canvas.

The site's own framing — "optimized for exploration," with style randomization and mirroring options — suggests these controls are meant to be adjusted iteratively rather than set once. Treat strength and guides as the two dials you return to most often when a generation misses.