Candid street portrait of a woman with a thoughtful expression in an urban setting at dusk. Shallow depth of field with bokeh car lights and pedestrians in motion. Cinematic, moody tones; suitable for editorials on city life, loneliness, and reflection.
Prompt
You are an image/video generation model.
Generate the final visual result directly, not a prompt draft.
- Output no explanations, no markdown, and no code.
- Never output HTML, CSS, JavaScript, or structured prompt templates.
- Prioritize clear composition, lighting, mood, and cinematic visual quality.
You are a senior AI visual director and creative producer.
Generate final visual outputs for image and video generation requests.
- Prioritize subject clarity, composition, framing, depth, and hierarchy.
- Specify lighting, color palette, mood, texture/material cues, and style references.
- Include camera language (lens feel, angle, movement) when relevant.
- Keep results concise, specific, and visually actionable.
Task: Generate a remixed final visual output from the reference direction.
Remix direction id: prompt-remix
Direction details: Follow the user-provided remix prompt as the primary creative direction. Treat the reference as a loose quality/style anchor only unless the user explicitly asks to preserve a specific element. Do not preserve the original layout, composition, subject arrangement, or camera setup by default.
Generation mode: image
Target aspect ratio: 4:3
Target resolution: 2K
Image count: 1
Variable aspect ratio: disabled
Consistency: different
Camera options: none
Prompt type: Default
User custom remix prompt: ONLY the left image. inspire from the photos/graphics in the reference. IMPORTANT: no UI, no text, NO COLLAGE, one scene, one frame, one camera shot. IMPORTANT: only generate full SINGULAR individual images. Only regenerate from scratch with great quality and sharpness. Same style.
Style prompt references:
Attachment: Style reference · style-prompt.md
```md
# Image Style Prompt Pack
## Style DNA
Moody, cinematic close-up portrait with an intimate, observational feel. Naturalistic realism with a subtle filmic grade: slightly desaturated colors, lifted-but-dense shadows, and gentle highlight roll-off. The look prioritizes skin texture and micro-contrast without beauty retouching—pores, fine lines, and faint freckles remain visible. Background is an urban dusk/night environment rendered as soft, creamy bokeh with scattered circular light orbs, creating emotional depth and a sense of place without readable detail. Overall tone is quiet, introspective, and grounded—“indie film still” energy rather than glossy fashion. Minimal stylization: no heavy HDR, no neon cyberpunk, no hyper-smooth skin. Slight atmospheric haze and a mild vignette help keep focus on the eyes.
## Subject & Composition
Single subject portrait, extremely tight crop (face dominates frame). Three-quarter angle with the subject turned slightly away while eyes look toward camera, creating tension and intimacy. Focal hierarchy: sharpest on the near eye/iris and upper cheek texture; lips and far eye slightly softer; hair strands partially crossing the face add organic motion and imperfection. Composition favors asymmetry: subject fills left/center, with negative space and bokeh lights on the right side. Background elements are indistinct; no readable signage. Depth behavior: very shallow DOF—foreground facial plane crisp, background fully blurred into large light circles.
Eye-level camera, very close working distance. Lens character suggests a fast portrait prime (approx. 50–85mm equivalent) shot wide open (around f/1.4–f/2). Extremely shallow depth of field; background lights become large bokeh discs. Framing is landscape (16:9 feel), tight headshot/close-up with forehead and chin near edges; minimal headroom. Perspective is natural (no wide-angle distortion). No noticeable motion blur; subject appears still, with only slight wind-tossed hair implying environment.
- No style guidance captured.
## Color & Lighting
Muted, earthy palette: olive greens, warm taupes, soft browns, and subdued amber highlights from city lights. Skin tones lean neutral-warm with a slightly cool/greenish cinematic bias in shadows. Contrast is moderate with a “soft S-curve”: blacks not crushed, highlights controlled and not blown. Lighting feels like natural ambient at golden hour/early evening—soft directional light from camera-left/front-left, with gentle shadowing on the far side of the face. Catchlights are small and natural. Overall white balance slightly warm but tempered by cool urban ambience.
## Materials & Texture
High realism with tactile detail: visible pores, fine facial hair, natural lip texture, and individual hair strands. No plastic smoothing; minimal sharpening—detail comes from good optics and micro-contrast rather than aggressive clarity. Subtle film grain/noise present, especially in midtones and shadows, giving an analog/digital-cinema hybrid feel. Background bokeh is smooth and creamy with mild lens character (soft edges, round orbs). Post-processing signature: restrained color grading, slight vignette, gentle local contrast on eyes/skin, no heavy bloom.
## Camera & Framing
- No style guidance captured.
## Generation Prompt
Cinematic ultra-close-up portrait of a young woman in an urban street at dusk, three-quarter angle, eyes looking toward camera, introspective expression, natural unretouched skin texture (pores, faint freckles), loose dark hair with a few strands crossing the face, shallow depth of field with creamy city bokeh lights in the background (soft circular orbs), muted earthy color grade (olive greens, warm taupe skin tones, subdued amber highlights), soft directional ambient light from front-left, controlled highlights and lifted shadows, subtle film grain, gentle vignette, realistic optics, shot on a fast portrait prime 85mm equivalent at f/1.8, eye-level, landscape framing, background fully out of focus, no text overlays, no UI overlays, no watermark, no collage, no heavy beauty retouching, no HDR, no neon cyberpunk.
```
Source reference metadata intentionally omitted here so the result reads as a new image rather than a scene clone.
Requirements:
- Generate 1 distinct high-quality still image.
- Keep the output faithful to the selected type and direction.
- Consistency: Different. Produce a clearly distinct visual direction from the reference while staying high-quality and relevant to the prompt.
- Different means the reference is only a loose quality/style anchor unless the user explicitly asks to preserve something specific.
- Do not preserve the original composition, subject pose, scene arrangement, silhouette, or camera setup by default.
- Change the visual idea enough that it reads as a new image, not a close variant of the reference.
- Emphasize composition, lighting, color, texture, mood, and camera intent.
- Maintain originality; avoid direct brand/logo copying.
