For this week, I created a mini project using p5.js and the Gemini API. The inspiration came from the “blind poem” （双盲诗）challenge popular on Chinese internet platforms, where participants create upper and lower lines of a poem without seeing each other’s contributions. I adapted this idea to an interactive experience between a user and AI. The project features two modes:

• User → AI: the user submits a question and the AI generates a poetic response.
• AI → User: the AI first generates the upper line of a poem, then the user inputs the lower line, forming a complete blind poem.

According to the role the AI plays, I designed two different prompts to make it generate either poetic responses or poetic questions.

Outcome

Demo

https://drive.google.com/file/d/1v2pPNPT4AgoaLlLSNMnMgMg7AGEpo3vc/view?usp=sharing

Limitation

One limitation of the current project is that it may not fully achieve the “blind” aspect of a blind poem. Although the AI is prompted to generate text independently of the user input, its outputs tend to have very similar styles. To improve the project, I could continue experimenting with the AI’s text generation to diversify its poetic style. Additionally, adding options for different text styles or moods could make the blind poem experience richer and more varied.

*Collab with ChatGPT