The wave arrives
TikTok refugees begin moving into rednote, landing in a product where every core surface is written in Chinese.
Design story behind #TikTokRefugees
In January 2025, millions of international users downloaded an app where nearly all the content they encountered was in Chinese. We had 48 hours to make it work for them — without disrupting the experience for the platform's existing daily active users.
RoleLead product designer (sole designer at launch)
Scope25 product surfaces · 8 teams
Collaboration18 designers · 5 PMs · 52 engineers
Companyrednote (Xiaohongshu)
TimelineJan 2025 – present
The moment
As the TikTok ban approached, rednote surged to #1 on the U.S. App Store. But every new user opened to the same experience: a wall of Chinese text they couldn't read.
The wave arrives
TikTok refugees begin moving into rednote, landing in a product where every core surface is written in Chinese.
rednote hits #1
The app climbs to the top of the U.S. App Store. The language gap is no longer a localization detail; it is the onboarding experience.
Manual translation ships
We move first with focused, manually translated surfaces so new users can navigate the product without waiting for a full system.
Auto-translation goes live
The work expands from selected copy to content comprehension, helping new users understand notes, comments, and community context.
Traffic peaks
The temporary response becomes infrastructure work: translation has to scale across surfaces, teams, and user expectations.
Text translation shipped in 48 hours, covering captions and comments. We could have continued with interface localization or new-user onboarding. But the community's content had to come first:
Could we make millions of notes, comments, and videos accessible to people who don't speak Chinese immediately?
The most practical response in that moment was to make auto-translation the default.
The auto-translate debate
Applying that default across all of rednote's UGC formats carried a serious risk: one bad translation could make it look like a creator had said something they hadn't. We had no playbook for doing this across so many formats, and I initially argued against it.
Then I stepped back from the individual creator perspective and considered the moment at platform scale. Translation quality could improve through iteration; the opportunity for millions of people to connect across a language barrier was immediate.
I changed my position and supported the default, while pushing for a minimum quality bar at launch and continued improvement afterward.
Auto-translation
We started with text, then extended auto-translation across every content format: images through OCR, videos with auto-generated subtitles, and livestreams with real-time captions. I worked directly with ML teams to improve image translation and voice dubbing quality — annotating cases, defining visual rules, and making aesthetic calls that fed back into the models.
The goal was to make translation work across every surface and content format.
Translation across every surface
Innovations born from real friction
Translation and subtitles aren't new. The harder problem was making them work the way people actually use rednote. The two solutions below redesigned both sides of the experience: how people view translated content and how creators decide who a note is for. Both later led to patent applications.
Consumption side
Draggable video subtitles
Friction
rednote's videos are layered — creators add their own subtitles, stickers, and text before posting. When we placed platform subtitles on top, they constantly collided with creator content. The viewing experience broke.
Exploration
Our first exploration was dynamic auto-positioning — using scene detection to place platform subtitles away from creator content. However, creator subtitles move constantly within the same video. Follow them, and our subtitles never stay still. Don't follow, and they overlap. Neither solved the problem.
Decision
I proposed anchoring subtitles to a default position relative to the bottom UI — consistent enough to read, without trying to predict where the content would land. Users who needed to reposition them could long-press and drag. The real fix wasn't a smarter algorithm. It was giving users control.
Creation side
Publish-side translation
Friction
Auto-translation made content readable. But readable isn't the same as intentional. Most content on rednote was written for a Chinese audience — international users could follow the words, but they weren't who the note was written for.
Exploration
One of our solutions was to give creators a way to translate their note before publishing — making the choice to speak to an international audience explicit. We first placed the action in the keyboard toolbar, assuming they might translate mid-composition. Testing showed a different pattern: people finished the note first, then decided who it was for.
Decision
We moved the translation action to the publish page — after writing, before publishing. Write in the language you think in. Translate when you're ready, then edit.
Leading at scale
Translation was only the beginning. The entire app — across 25 product surfaces and 8 teams — needed to go global. The biggest challenge for me, as a solo designer, was to get everyone aligned on work that was critical but outside anyone's core mandate.
Phase 0 · Jan (2 days)
As the sole designer, I drove the cross-language experience across core consumption surfaces — from concept to launch in 48 hours.
Phase 1 · Feb–Mar
Led 18 product designers across business units through a full-app multilingual UI sprint. Every surface, every language, one standard.
Phase 2 · Apr onward
Led field research in Singapore and Malaysia, then partnered with core product teams to redesign search, voice messaging, and content creation around international users' expectations.
Ongoing
Expanded the work from app to web and built globalization into the design system and accessibility standards so it could serve as a foundation rather than a patch.
How I scaled the work
From initiative to alignment
Scaling this work meant getting teams aligned around something outside their core mandate. I mapped every globalization failure across the product, turned it into a design guide, and brought it to the entire design team with a concrete proposal — the solution and the plan.
That became the shared standard 18 designers used to move the work forward across teams.
Created a living standard
I didn't want this to be a one-time sprint. So I wrote two docs — a design spec and an SOP — distilled from auditing 1,000+ multilingual issues across the app. They ended up shared company-wide, and are still referenced by design and product teams today.
Built credibility inside and out
I represented rednote's globalization design at two industry sessions: one with Apple designers at Shanghai Jiaotong University and one with a visiting delegation from Milan. Sharing the work externally also strengthened its credibility inside the company.
Impact
In the week of the TikTok Refugees event alone, rednote saw 3.84 million overseas downloads. The work continued after the traffic spike as a platform-wide globalization effort.
@Libbe De Witte 星婷✨A U.S. user posted on rednote, February 7, 2026
“Thank you for making it so easy for American users to use your app through the translation feature. And thank you to all the Chinese sisters (and brothers!) who welcomed us, shared memes, and opened their hearts. This little cultural exchange has made me a more worldly person.”
Reflection
This project pushed me well beyond my defined role. Alongside my design work, I took on broader product responsibilities during iteration, led one-on-one interviews and focus groups during field research in Singapore and Malaysia, and partnered with ML teams to define experience standards for image translation and voice dubbing.
The best cross-cultural design doesn't just translate words — it translates belonging.