Image to hook to draft

attention

Turn an image into its strongest visual hook, the question viewers are most likely to ask, and a usable copy draft. attention is built for creators who post image-led content and developers who want that same intent -> copy flow inside their own tools.

中文说明：先找到图片里最值得展开的注意力点，再生成一版可继续修改的中文文案草稿。

Open the GitHub repo Use via API / MCP
attention demo screenshot

For creators

Use the browser demo to upload an image, bring your own model key, and get a draft you can keep editing.

  • Works well for fashion, accessories, cafe details, and mood-led posts.
  • Good when you know the image is strong but the caption is still vague.

For developers

Integrate the same flow through CLI, HTTP API, MCP, or the packaged skill.

  • Shared schema: attention.v1
  • Best fit for creator tools, extensions, and agent workflows.

Try locally in under a minute

Until a dedicated hosted demo is wired up, the fastest reliable path is the local browser demo.

  • python3 -m pip install -r requirements.txt
  • python3 app.py --inbrowser
  • Upload an image and bring your own provider key.

Run one developer flow quickly

  • python3 -m pip install -e .
  • attention-api --host 127.0.0.1 --port 8000
  • python3 scripts/http_demo.py --image /absolute/path/to/image.jpg --provider gemini --api-key "$GEMINI_API_KEY"

Example image patterns

Start with the kind of image you care about, then see what `attention` should try to extract.

Repo: rickqdev/attention