Ирина БороганPromptUsing the provided photo of a female model wearing a light beige chunky knit zip-up cardigan with a high stand collar, perform the following actions precisely and preserve natural appearance:
1. Color correction: remove overall yellow cast, neutralize skin tones while keeping warmth, increase vibrancy and saturation for a bright, rich look, increase contrast moderately and slightly lift midtones for an editorial, vibrant fashion aesthetic.
2. Detail enhancement: reduce noise, increase sharpness on facial features and fabric texture, enhance metallic highlights on small gold earrings without overexposure.
3. Composition: crop and reframe to vertical 9:16 aspect ratio suitable for Instagram Reels and TikTok, keep natural headroom and full view of the upper garment, maintain original pose and perspective.
4. Layer exports and masks: produce high-resolution color-corrected master image and create separate high-precision alpha masks (PNG) for: (A) model silhouette, (B) hair, (C) upper garment (cardigan) edges, (D) background. Masks should be tightly fitted with slight feathering (2–4 px) to avoid hard edges.
5. Background variant: produce an alternate version where the background is subtly desaturated and given a gentle Gaussian-style blur (simulating depth-of-field) while preserving the contrast and detail on the model and clothing.
6. Constraints: do not alter facial features, do not change clothing design, do not remove accessories. Keep lighting consistent with original scene.
Deliverables: a) color-corrected, framed high-res image (PNG/JPEG); b) alternate background-blurred version; c) four separate alpha masks (model, hair, garment, background) sized to the final 9:16 crop and aligned to the corrected images. Prepare these assets for subsequent upscaling and animation (no animation yet). Provide filenames and metadata indicating recommended output resolution for final export.
Style note: aim for a crisp, editorial fashion look optimized for mobile vertical video — bright, saturated, modern, clean.