1940s B&W High Contrast · عوامی Open Skill
1940s B&W High Contrast
1940s B&W High Contrast Digen پر ایک عوامی Open Skill ہے۔ کھولیں اور دیکھیں یہ ورک فلو کیا تخلیق کرتا ہے۔
Beauty
0 0 0
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4150474/2026-10-15/qqa4rj2m.jpg
# Film Noir Portrait (1940s B&W High-Contrast)
Restyles a user-supplied image into a 1940s Hollywood film-noir still: neutral black & white, hard chiaroscuro, crisp venetian-blind light bands raking across the body, and half of the face sunk into darkness.
This is a single-step image-to-image skill. There is no project state, no asset collections to update, and no config files — the identity, pose and framing come from the user's uploaded image; everything else comes from the style spec in `references/`.
## When to use
- The user uploads (or points at) an image and asks for a 40s noir / high-contrast black-and-white / venetian-blind-shadow / half-face-in-darkness look.
- The user asks to push an already produced noir frame further ("darker", "more blinds", "hide the other side of the face", "add smoke", "more 40s glamour").
- The user describes the scene without naming it: "old detective movie lighting", "black-and-white with window blind shadows".
## Before composing anything
Read `references/film-noir-style.md` once per conversation, using `read_skill_file`. It holds the layered style spec (base look, lighting, blind stripes, subject styling, environment/props, composition, finish), the exact English prompt template, the avoid-list, and the variation knobs. Do **not** improvise the look from this file alone — the reference is the contract.
## Steps
1. **Identify the source image.** Use the `asset_id` of the image attached in this turn (or the one the user pointed at). If several images were attached and the target is unclear, ask which one to use — one short question, nothing more.
2. **Read the source (recommended).** Call `recognize_image` on the source to lock down what must be preserved and what must change:
`recognize_image(image_url=<source asset_id>, prompt="Describe in English: subject and identity cues, pose and head angle, framing and crop, existing light direction, wardrobe and hair, background elements. Note which side of the face is currently lit.")`
Use the answer to decide: which side of the face stays lit (mirror the existing key light), which side falls into shadow, where the blind bands should cross the body (forearm, chest, hip — the guide requires them on the subject, not only on the wall), and whether the original background should be replaced by the period room.
3. **Compose the prompt.** Follow the layer order and the fill-in template in `references/film-noir-style.md`. English only, roughly 120–220 words, no bullet lists inside the prompt. Always state: preserve the subject's face, hair silhouette, pose and crop from the reference image; convert to pure neutral black & white; keep one half of the face in deep shadow; hard-edged diagonal venetian-blind light bands across the subject's body; 1940s period room and props; fine film grain and crushed blacks.
4. **Generate:**
`generate_image(input_urls=[<source asset_id>], prompt=<composed prompt>, model='i2i', orientation=<match the source; default 'portrait' for people>, name='Noir — <3-4 word label>', entity_type='film_noir_portraits')`
Never pass a physical model name; the stable channel here is always `i2i` (bare channel — do not use `.hd*` variants, so that all subjects remain renderable).
5. **Variations.** `batch_size` is fixed at 1, so alternates mean another call: same source image, same prompt skeleton, change exactly one knob from the variation list in the reference (blind angle/density, shadow side, prop emphasis, framing). Tell the user which knob you turned.
6. **Reply.** Two or three natural-language sentences describing what was produced (look, light, mood). Do not paste URLs and do not dump the full prompt unless asked. When the user is likely to want a next step, call `set_guidance` in the same turn (e.g. "deepen the shadow on the other half of the face", "switch to a vertical half-body framing", "try a different venetian-blind angle").
## Parameter conventions
| Parameter | Convention |
|---|---|
| `model` | alw