Good Person Card · skill publik Open Skill

Good Person Card

Good Person Card

@ExploreFly

Internal

0 0 0

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/3656543/2026-10-14/2cx6pdqk.jpg

# Good Person Card

Turn the user’s uploaded photo into a **warm, comforting** version of the same scene, with **one short positive or self-motivational line** integrated into the image (readable typography, not a separate caption in chat only).

There is no structured project state—progress lives in the conversation and canvas assets.

## When to use

- User uploads a photo and wants it to feel warmer, softer, or more “cozy.”
- User wants **encouraging / motivational / wholesome text on the image** (good-person card, kindness note, daily affirmation).
- User asks to redo the quote, mood, or model quality for a card they already started.

## Inputs you need

| Input | Required | Notes |
|-------|----------|--------|
| Reference image | **Yes** | User attachment or an `asset_id` / URL from this chat. |
| Quote text | Preferred | If missing, propose 2–3 options (see `references/quotes.md`) and use the one they pick—or the best fit if they say “surprise me.” |
| Quote language | Optional | Default: **English** for on-image text and chat unless the user explicitly asks for another language. |
| Orientation | Optional | Match the reference (portrait / landscape / square). Use `recognize_image` only when aspect is unclear. |

## Model selection (image-to-image only)

Use **`generate_image` with `input_urls=[reference]`**. Allowed channels—**do not use bare `i2i` or physical model names**:

| Channel | When to use |
|---------|-------------|
| `i2i.hd.lite` | Default for a quick warm card or first draft. |
| `i2i.hd` | User wants higher visual quality, sharper details, or retries after a lite result they dislike. |
| `i2i.hd.gpt` | User asks for the “best” text on image, maximum quote legibility, or explicitly names GPT / smart i2i. |

If the user does not specify quality, start with **`i2i.hd.lite`**. Escalate to `i2i.hd` or `i2i.hd.gpt` on request or after a failed legibility retry.

`list_models` is for troubleshooting only—do not pick channels from memory outside the table above.

## Steps

1. **Confirm the reference**  
   - If no image is attached, ask the user to upload one before generating.

2. **Quote**  
   - If they did not provide text, call `read_skill_file(good-person-card, references/quotes.md)` and offer a few choices aligned with their mood (gentle vs bold motivation).  
   - Keep quotes **short** (roughly 4–18 words) so they fit on the image.

3. **Optional: understand the photo**  
   - When composition matters (faces, pets, landscapes), call `recognize_image` on the reference to avoid covering key subjects with text.

4. **Generate the warm card (single output)**  
   - Call `generate_image` with:
     - `model`: one of `i2i.hd.lite`, `i2i.hd`, `i2i.hd.gpt` (see table).
     - `input_urls`: `[reference]` (asset_id preferred when available).
     - `orientation`: `portrait` | `landscape` | `square` to match the reference when known; omit only if unknown (platform default applies).
     - `prompt`: English (see template below)—embed the **exact on-image quote** (English by default).
     - `name`: short canvas title, e.g. `Warm card` or first few words of the quote.
     - `entity_type`: `warm_cards`

5. **Reply**  
   - Summarize mood + quote in **English**. **Do not paste long raw URLs**—the asset appears on the canvas.

6. **Iterations**  
   - New quote or warmer look → regenerate with the **same reference** (or the last good output if they prefer that as the new reference—confirm if ambiguous).  
   - “Clearer text” → retry with `i2i.hd.gpt` or `i2i.hd` and tighten the typography lines in the prompt.

## Prompt template

Use one prompt per call; replace `{QUOTE}` and tune placement:

```text
Image-to-image edit of the reference photo. Keep the same main subject, identity, and overall composition. Transform the mood to warm, cozy, and comforting: soft golden-hour lighting, gentle warm color grading, subtle soft glow, inviting atmosphere. Do not add new people or change the core scene into something unrelated.