Trained on Civit with Hunyuan Video ! Retrain of <a target="_blank" rel="ugc" href="https://civarchive.com/models/1473927/djz-westwick-detective-elijah">Westwick Detective, Flux lora</a> (narrative illustration set) Style used with this Set: <a target="_blank" rel="ugc" href="https://civarchive.com/models/1478166/djz-westwick-style-hyvid">https://civarchive.com/models/1478166/djz-westwick-style-hyvid</a> Executive SummaryThis analysis examines a collection of prompts used to train a generative image model focused on creating detective/noir aesthetic images. The model appears to be designed to generate scenes featuring a male character in various moody, atmospheric urban environments with consistent visual elements including wet reflective surfaces, industrial settings, and distinctive lighting.<h2 id="common-phrases-and-trigger-words-jky18wz02">Common Phrases and Trigger Words</h2><h3 id="primary-trigger-phrase-ep0vtd07j">Primary Trigger Phrase</h3><ul><li>"photo of kgk man, detective aesthetic" - This appears in every single prompt and seems to be the core trigger for the model.<ul><li>"kgk" appears to be a specific identifier code for the character model</li><li>"detective aesthetic" establishes the visual style/genre for the images</li></ul></li></ul><h3 id="environmental-elements-dsf8c4g1n">Environmental Elements</h3><ul><li>"wet, reflective surface"/"puddle" - Appears in 12/25 prompts, indicating this is a core visual element</li><li>"foggy"/"fog" - Appears in 7/25 prompts</li><li>"rainy"/"rain" - Appears in 5/25 prompts</li><li>"flooded" - Appears in 3/25 prompts</li></ul><h3 id="lighting-and-atmosphere-22lm1rr8v">Lighting and Atmosphere</h3><ul><li>"red lights"/"red lighting" - Appears in 11/25 prompts</li><li>"blue lights"/"blue hues" - Appears in 4/25 prompts</li><li>"moody"/"mood" - Appears in 20/25 prompts</li><li>"night"/"nighttime" - Appears in 5/25 prompts</li><li>"dark" - Appears in 6/25 prompts</li><li>"atmospheric" - Appears in 5/25 prompts</li><li>"dimly lit" - Appears in 5/25 prompts</li></ul><h3 id="character-descriptions-izuekukty">Character Descriptions</h3><ul><li>"wearing a long coat"/"trench coat" - Appears in 14/25 prompts</li><li>"hat" - Appears in 7/25 prompts</li><li>"standing" - Appears in 24/25 prompts (the only exception mentions "walking")</li><li>"facing away from the camera" - Appears in 5/25 prompts</li></ul><h3 id="settings-vazdes2vg">Settings</h3><ul><li>"industrial" - Appears in 8/25 prompts</li><li>"urban" - Appears in 5/25 prompts</li><li>"rooftop" - Appears in 3/25 prompts</li><li>"street" - Appears in 7/25 prompts</li><li>"building"/"buildings" - Appears in 12/25 prompts</li><li>"dock"/"dockside"/"waterfront" - Appears in 3/25 prompts</li><li>"cathedral" - Appears in 2/25 prompts</li></ul><h2 id="image-model-capabilities-49oq3req7">Image Model Capabilities</h2>Based on the analysis of these prompts, the image generation model appears capable of producing:<ol><li>Character Portrayal:<ul><li>A consistent male character in detective-style attire (long coat, sometimes hat)</li><li>Character is typically shown in contemplative poses, often with back to camera</li><li>Limited front-facing or detailed facial depictions</li></ul></li><li>Environmental Settings:<ul><li>Urban nightscapes with distinctive architecture</li><li>Industrial areas with machinery and structures</li><li>Rooftops overlooking cities</li><li>Water-adjacent locations (docks, waterfronts)</li><li>Abandoned or desolate buildings (including cathedrals)</li><li>Interior spaces with specific architectural elements (high ceilings, large windows)</li></ul></li><li>Atmospheric Effects:<ul><li>Wet surfaces with reflections</li><li>Fog and rain</li><li>Specific lighting conditions (especially red and blue lighting)</li><li>Night scenes with urban illumination</li><li>Flooded areas</li></ul></li><li>Emotional Tone:<ul><li>Consistently moody, somber, contemplative</li><li>Mysterious, eerie, atmospheric</li><li>Isolated, desolate</li><li>Introspective</li></ul></li></ol><h2 id="prompt-structure-analysis-x8mr6ozi8">Prompt Structure Analysis</h2>The prompts follow a consistent structure:<ol><li>Opening trigger phrase: "photo of kgk man, detective aesthetic"</li><li>A period followed by descriptive text: ". The video depicts..."</li><li>Detailed scene description focusing on:<ul><li>Character positioning and appearance</li><li>Environmental setting</li><li>Lighting and atmospheric conditions</li><li>Mood and emotional tone</li></ul></li></ol>Notably, all prompts refer to "The video depicts" despite appearing to describe still images. This suggests the model may have been trained on frames extracted from video content or is intended to generate images that have a cinematic quality.<h2 id="limitations-and-scope-ort3msr6i">Limitations and Scope</h2>The model appears to be narrowly focused on generating noir/detective-style scenes with a specific male character in urban environments. There is no evidence in these prompts of:<ul><li>Daytime scenes (all appear to be night or dimly lit)</li><li>Rural or natural settings</li><li>Multiple characters or social interactions</li><li>Action sequences (character is almost always standing still)</li><li>Close-up portraits or detailed facial expressions</li><li>Bright or cheerful environments</li><li>Modern or futuristic settings (most have an industrial or historical feel)</li></ul><h2 id="conclusion-uh7j4irqp">Conclusion</h2>This image generation model has been trained to produce atmospheric noir/detective scenes featuring a specific male character in moody urban environments. The consistent trigger phrase "photo of kgk man, detective aesthetic" appears to be essential for accessing the model's capabilities. The training data emphasizes wet reflective surfaces, fog, industrial settings, and distinctive red/blue lighting to create a consistent aesthetic across generated images.The model's scope is intentionally narrow, focusing on creating a specific mood and visual style rather than offering versatility across different genres or settings. This suggests it may be a character-specific model designed for a particular creative project with a consistent visual identity.

Files with SHA256: abf984cb36aab241f111c5df358c759ea7e49ad1a97eee9304050b377229cc16