Shangtouwa:AI Interactive Storytelling Community

Make AI creation a journey users can shape.

Shangtouwa is an AI interactive storytelling community where readers can become participants and creators. The product challenge was to connect discovery, story interaction, creation, and reusable settings into a coherent experience.

I am Wang Sen, an undergraduate at Beijing University of Posts and Telecommunications, graduating in 2027. During my May–September 2026 internship at Tencent PCG, I worked on AI application product design and iteration: needs analysis, product decisions, Figma prototypes, interaction design, reviews, engineering alignment, launch follow-up, and quality checks.

I created the Figma designs for the work presented here. My role connected product intent with implementation through collaboration with engineering.

From a rough idea to a work ready to publish.

Some creators arrive with a detailed setting; others need help getting started. I designed free-form input and step-by-step guidance, connecting world and character setup and reuse with AI inspiration editing and confirmation, then generation and publishing.

1. PrepareFree-form input or guided setup; create or reuse worlds and characters.
2. ShapeEdit and confirm AI-generated inspiration before proceeding.
3. PublishGenerate and publish interactive novels, vertical comics, or short narrative scenes.

These are three content formats within one product, supported by two reusable asset types: worlds and characters. I completed the publisher, creation-assistant, and inspiration-editing designs, participated in engineering collaboration, and helped bring the features to launch.

Product judgment beyond the creation screen

Trial feedback and page walkthroughs informed a separate set of improvement proposals. I prioritized them by user value, reach, and implementation cost.

• Respect hard boundaries. I distinguished content users need to block from ordinary preferences that should lead to fewer recommendations.
• Protect choices before expanding management. To address branch reselection overwriting progress, I proposed a staged path: confirm and save separately, then a branch list, then a timeline.
• Define reuse through user intent. I helped align asset-reference counting around whether a user actively selected an existing world or character during input. Model-generated content does not count as user reuse.

Meet moderation requirements without losing creative intent.

A custom style requires a reference image and a name. Both need moderation before the style can be saved and used. The design challenge was to make those checks explicit while preserving the creator's work when something failed.

I designed the reference-image upload, image and name moderation, saving, and selection flows in Figma. I participated in cross-platform reviews, aligned behavior with engineering, and followed the feature through launch.

• Prevent premature or duplicate submissions. Disable the relevant buttons while moderation or saving is in progress.
• Keep useful input after rejection. If the image fails moderation, retain the name and allow a new upload. If the name fails, retain both the image and name so the user can edit it.
• Make technical failures recoverable in place. Keep the user in the current flow after an API failure and allow a retry.
• Complete the selection for the user. After a successful save, close the overlay, automatically select the new style, and scroll it into view.

The feature is live. For validation handoff, I also helped prepare 85 functional acceptance-test cases covering normal and exception paths, including moderation, drafts, derivative creation, and account isolation.

Connect free rewards with paid energy top-ups.

I designed the energy center and worked with engineering to bring it live. It brings together four ways to obtain energy: rewarded videos, daily check-ins, invitations, and paid top-ups.

I placed a rewarded-video entry point in the generation waiting period and connected reward crediting with the balance and transaction history. The flow makes the next step visible, from choosing how to obtain energy to checking that it has arrived.

I also participated in campaign banner reviews and launch coordination. The banners are live. This work connected the free incentive and paid replenishment journeys through product and interaction design.

Turn product rules into checks the team can act on.

I translated issues such as character inconsistency, asset reuse, characters appearing despite a “skip character” choice, and failed or stalled images into concrete checks.

I organized 96 scenario designs across four world categories, six character categories, and four plot-intent categories. I also documented 82 historical case-level observations of generation time and exceptions to support review and follow-up checks.

Using Codex and scripts, I connected screenshot evidence capture, result recording, and report compilation. The workflow supported human review and targeted retesting of key issues.

What I brought to delivery

My contribution spanned the publisher, creation assistant, inspiration editor, custom styles, and energy center: turning needs into prototypes and interaction rules, discussing implementation in reviews, following launch progress, and checking quality. Recommendation and branch-management improvements remain presented here as proposals.