kie-gpt-image

Generate and edit images with **OpenAI GPT Image 2** through the Kie.ai jobs API. Best for photorealism, accurate text-in-image, product shots, and clean edits. > **Setup:** needs `KIE_API_KEY` in `~/.claude/.env`. ## Modes | Mode | Model id | Refs | Notes | |---|---|---|---| | text-to-image (default) | `gpt-image-2-text-to-image` | — | Pure prompt → image | | image-to-image / edit | `gpt-image-2-image-to-image` | up to 16 | Auto-selected when any `--image` is passed | ## Flow 1. **Parse** `$ARGUMENTS`. If the user supplies image(s) → image-to-image mode (auto-detected). 2. **Settings** (ask, or defaults): - **Aspect ratio:** `auto` (default). Options: `1:1`, `3:2`, `2:3`, `4:3`, `3:4`, `5:4`, `4:5`, `16:9`, `9:16`, `2:1`, `1:2`, `3:1`, `1:3`, `21:9`, `9:21`. - **Resolution:** `1K`, `2K` (default), `4K`. - **Variants:** default 3. 3. **Craft the prompt.** Show before generating. For edits, say what to keep vs change. 4. **Generate:** ``` python3 ~/.claude/skills/kie-gpt-image/generate.py "<prompt>" \ --aspect-ratio auto --resolution 2K --count 3 \ [--image <url-or-path> ...] # any --image switches to image-to-image (up to 16) --slug <slug> ``` 5. **Present:** output → `~/images/gpt-image/<date>-<slug>/`. Offer more variants, 4K, or an edit pass on a favorite. ## Resolution constraints (enforced by the script) - `5:4` and `4:5` → **1K only**. - `1:1` cannot go **4K** (falls back to 2K). - `auto` aspect ratio is limited to **1K** — set an explicit ratio for 2K/4K. ## Notes - Output is PNG. Local `--image` files auto-upload via the shared `_kie` client. - For YouTube thumbnails specifically, `/yt-thumbnail` is still the tuned tool; this is general-purpose.