Beach Swimwear Duo Photo · skill Open Skill công khai
Beach Swimwear Duo Photo
Beach Swimwear Duo Photo là một Open Skill công khai trên Digen. Mở ra để xem workflow này tạo gì.
Custom
3 0 0
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/3656543/2026-10-14/pxhpe5fs.jpg
# Beach Swimwear Duo Photo Turn the user’s uploaded reference photo into a **bright coastal beach** scene featuring **two people in summer swimwear**, composed as a **natural duo portrait** (side-by-side or lightly interacting). Identity and overall likeness should follow the reference via **image-to-image (I2I)** only—do not use text-to-image for the final result. ## When to use - User uploads one or more photos and asks for a **beach / seaside / sandbox** vibe with **swimwear** and **two people together**. - User wants to **regenerate**, **change swimwear color**, **adjust mood** (sunset vs midday), or **switch I2I quality tier** while keeping the same upload. ## Reference file (read before generating) | Purpose | File | |--------|------| | Beach setting, swimwear styling, poses, lighting, and prompt templates | `references/beach_swimwear_i2i_guide.md` | After `read_skill`, call **`read_skill_file(beach-swimwear-duo-photo, references/beach_swimwear_i2i_guide.md)`** once per session (or again if the user changes style constraints). ## Workflow 1. **Inputs** - Require at least **one reference image** from the user (attachment or canvas asset). Prefer a photo that clearly shows **two people**; if only one person is visible, ask whether to duplicate styling for a second figure or wait for a duo photo—do not invent a second face without user consent. - If multiple uploads exist, pick the clearest duo shot; note others only if the user asks to blend traits. 2. **Understand the reference** — `recognize_image` - Use the user’s image URL / asset id. - Confirm: number of visible people, approximate ages (adult-only output), genders presentation if relevant to swimwear cut, skin tone, hair, and **must-keep likeness** cues. - Summarize briefly for the user in their language; keep **generation prompts in English**. 3. **Plan the look** (align with user; defaults if silent) - **Setting:** tropical or classic sandy beach, ocean horizon, soft waves (see reference guide). - **Wardrobe:** tasteful **summer swimwear** for two (e.g., one-piece / bikini + board shorts / swim trunks)—**no nudity, no explicit content** (required for `i2i.hd*` channels). - **Framing:** `orientation` from reference when obvious; otherwise `portrait` for waist-up duo selfies, `landscape` for full-body walking shots. Use `square` only if the user asks. - **Resolution:** omit `resolution` unless the user requests higher quality (then prefer `1080P` on `i2i.hd` / `i2i.hd.gpt`). 4. **I2I model selection** (hard-coded channels only) | Channel | When to use | |---------|-------------| | `i2i.hd.lite` | Fast preview, lower cost, user wants a quick draft | | `i2i.hd` | **Default** balanced quality for the final duo beach photo | | `i2i.hd.gpt` | User asks for maximum fidelity / “best quality” / GPT-backed I2I | Do **not** use physical model names. Optional: `list_models` with `model_type='image'` only if the user reports a channel error. 5. **Generate** — `generate_image` ```text input_urls: [reference image — asset_id, https, or s3:// from the upload] model: one of i2i.hd.lite | i2i.hd | i2i.hd.gpt prompt: English, built from templates in references/beach_swimwear_i2i_guide.md name: short canvas title, e.g. "Beach swimwear duo" entity_type: skill_generation orientation: portrait | landscape | square (see step 3) ``` - Always pass **exactly one** primary reference in `input_urls` unless the user explicitly supplies two references for two identities. - Emphasize in the prompt: **preserve faces and body proportions**, **same two individuals**, **new beach background and swimwear outfits**. 6. **After delivery** - Describe what changed (setting, outfits, lighting) in natural language; **do not paste long URLs**. - Offer `set_guidance` options: alternate model tier, sunset vs blue-sky, different swimwear colors, closer vs wider framing. - On failure: report the tool error honestl