Added Oct 28, 2026
Sofa Travel
Stay on the sofa and let an AI "take" you to Paris or Kyoto: it plans a set of shots per destination and hands you Chinese image prompts for Doubao or Jimeng, or generates the album itself.
- Ecosystem
- Agent Skills
- Skill category
- Media
- Creator / organization
- HeiGeAi
What it is
Sofa Travel is a Chinese-language skill by HeiGeAi. The README's line is that your album tours the world before you leave the house. You give one photo of yourself, or choose faceless mode, say where you'd like to go, and the agent plans a set of shots for that destination: arriving, wandering, taking a break. The default is three vertical photos in the style of a friend's casual phone snaps, extendable to six or nine. An agent with image generation makes the pictures and files them into a local album; one without hands you the full Chinese prompts, reference-photo notes, retouching instructions and share captions to paste into Doubao or Jimeng. Twelve destinations are built in, each with landmark, street, table, scenery and evening scenes, and there are modes for two people, pets and custom places. What sets it apart is faceless mode and its honesty rules: the skill says not to claim it booked flights, and when a tool fails it keeps the finished pictures and prompts, and does not pass off stock images as results.
Who it's for:Anyone who wants a set of "pretend trip" photos of themselves, a friend or a pet without working out the prompts. If you can use Doubao or Jimeng, you can use this.
Before you start: It's built for file-reading agents such as Codex, Claude Code and WorkBuddy, and can run with no install at all by having the agent read START.md and SKILL.md from the repo; the command-line path needs Python 3.10 or newer with no extra packages. Everything is in Chinese. The README says the local playbook page works offline and takes no photos; when you do generate images, your photo goes to whichever image tool you use, under that tool's rules. Code and original prompts are free under the MIT license; third-party image generation costs or free quotas depend on the tool. The sample people are fictional, and results and likeness vary by platform.
Based on the project's public documentation. Not tested by AI Inspiration World.