On Creating My Styles

If you've ever typed "in the style of Van Gogh" into an AI image generator, you've probably noticed the results can be... generic. You get swirly skies and heavy brushwork, but it feels like a costume rather than a transformation. I developed what I call "triangulated style fusion" to solve this problem—creating style prompts that are both technically precise and universally applicable. The Core Idea Instead of naming a single artist, I triangulate three reference points: two artists and one cultural or atmospheric element. This creates a more specific and reproducible style direction for the AI, while avoiding the shallow interpretations that single-artist prompts often produce. The weighting matters: roughly 50% from the primary artist, 30% from the secondary, and 20% from the cultural element. This ensures each contributor has a defined role rather than muddying into an average. The Seven-Step Process Step 1: Identify Core Visual Techniques I start by identifying 2-3 visual techniques that could enhance any image—color temperature, brushwork character, light treatment, edge behavior. The key constraint: these must be universal. If I catch myself writing "enhances skin tones," I've already failed, because that assumes a portrait. The techniques need to work equally well on a face, a seascape, an abstract, or a coffee cup. Step 2: Extract the Emotional Essence Next, I identify the universal energy or mood that can be amplified regardless of content. This isn't narrative mood ("mysterious" or "romantic") but pure visual energy—tension, tranquility, dynamism, weight. Step 3: Distill the Cultural Element The third element—a cultural, historical, or atmospheric reference—gets distilled to its pure visual properties. This is where precision matters most. Take "Film Noir" as an example. The content associations (detectives, femme fatales, rain-slicked streets) are useless to me. What I extract instead: high-contrast chiaroscuro, dramatic shadow angles, tension between illuminated and obscured areas, cool-shifted blacks. Those are visual methods, not subject matter. Or consider "Japanese Wabi-Sabi." I'm not thinking about tea bowls and weathered wood. I'm extracting: asymmetrical balance, muted earth tones, textural imperfection as beauty, restrained negative space. Step 4: Verify AI Executability Some techniques sound great but don't translate. Pollock's drip paintings are too chaotic for reliable AI reproduction. Certain conceptual approaches don't have clear visual correlates. I test whether the AI can actually execute what I'm describing. Step 5: Select Artists by Technique, Not Reputation Artist selection focuses on documented mastery of specific universal techniques—not fame, not personal preference. I'm asking: which artists have clear, reproducible methods that serve this fusion? Step 6: Triangulate the Fusion With all three elements decomposed, I articulate how they interact—not just coexist. The prompt structure makes relationships explicit: "where [technique] creates [effect] through [method]..." Step 7: Verify Universality Finally, I confirm the prompt would work equally well on a portrait, a landscape, an abstract composition, or a still life. If the style only makes sense for certain subjects, it's not truly universal. Decomposing Artists into Technique This is really the heart of the method. When I analyze an artist, I'm not asking "what does their work look like?" but rather "what methods produce their distinctive look?" I examine four attributes: Color behavior : How do they handle palette? Complementary tension or analogous harmony? Unexpected accents? What's their saturation tendency—muted, vibrant, selective? Light treatment : Do they create luminosity from within (think Rembrandt's glow), flatten it (Matisse), or fragment it (the Impressionists)? How do they handle transitions between light and shadow? Surface and texture quality : Visible brushwork or smooth blending? Thick impasto or thin glazes? Granular or fluid? Edge behavior : Hard boundaries or soft dissolution? Lost and found edges? The goal is isolating 2-3 techniques that are both distinctive to that artist and transferable to any subject matter. A Concrete Example Say I'm considering Sorolla. The generic approach would note: "Spanish, beach scenes, impressionistic light." My technique decomposition looks different. I see wet-on-wet brushwork that captures light as movement rather than static illumination. Cool shadow temperatures against warm highlights creating vibrant optical contrast. Visible directional strokes that follow form and suggest rather than define. High-key luminosity where even shadows retain transparency. Now I have methods rather than associations. The Second Artist's Role The second artist isn't just "another style to add"—they're chosen specifically to complement or productively tension with the first. I ask myself: What does Artist 1 lack that would strengthen the fusion? Where might Artist 1's techniques become monotonous without counterbalance? What universal technique does Artist 2 bring that Artist 1 doesn't? If Sorolla brings luminous fluidity and warm/cool contrast, maybe I pair with someone who brings structural weight (Cézanne) or atmospheric mystery (Whistler). The 30% weighting means the second artist modifies rather than competes. Manifesting the Fusion The final prompt articulates how the three elements interact: First, what Artist 1's technique does to the overall image. Second, how Artist 2's technique modifies or layers onto that. Third, how the cultural element infuses atmospheric quality throughout. This way, the AI isn't just averaging three references—it's applying a specific technical recipe. My Quick Validation Test If I can't complete this sentence, my decomposition isn't precise enough: "Artist X's contribution manifests as [specific visual technique] which affects [color/light/texture/edge] by [specific method], regardless of whether the image contains a face, a tree, or an abstract shape." That "regardless of" clause is the universality check. If the technique only makes sense for portraits or only for landscapes, I go back and decompose further until I find the underlying method that truly transfers. The result is a style prompt that gives AI image generators something they can actually execute—not a vague aesthetic reference, but a technical specification that transforms any source image through precise, reproducible methods. To bring this full circle, here's what a complete fusion prompt looks like using the Sorolla example from above: Sorolla + Whistler + Nordic twilight atmosphere, where Sorolla's wet-on-wet brushwork creates luminous movement through cool shadow temperatures against warm highlights, Whistler's tonal restraint produces atmospheric depth via soft edge dissolution and limited palette harmony, and Nordic twilight infuses contemplative stillness through extended tonal gradients. The fusion manifests as directional strokes that follow form while dissolving into atmosphere, transparent shadows retaining inner light, and muted warmth emerging from cool veils. Sorolla brings luminous fluidity and warm/cool temperature contrast. Whistler counterbalances that vibrancy with tonal restraint and atmospheric softness—preventing the result from becoming too sunny or literal. Nordic twilight adds contemplative weight and those long, quiet tonal transitions that give the style emotional depth without specifying any content. The tension between Sorolla's Mediterranean brilliance and the cooler, more restrained qualities creates something neither artist would produce alone: The prompt in seedream The prompt in flux pro 1.1 Flux pushes the prompt towards photorealistic. Seedream executes the style pretty faithfully. I'm curious. How do you build your prompts?