Unique Research · Insight / Unique Friend Q&A · January 12, 2024 · Guest: Cao Renyi (Roy), Co-founder & Product Director, LibAI Lab
Inventory type remains insight (Unique Friend series — not partner welcome).
Guest
Cao Renyi (Roy) — Zhejiang University 2002; STEAM advocate; serial founder backed by IDG, Huachuang, Zhenshun. Co-founder & product director of LibAI Lab — leads Cutout.Pro and PromeAI.com. LibAI aims to fuse Eastern art peaks with global tech (“Li Bai”); 150M+ yearly active users across 70+ countries; a16z traffic ranks among global top-20 GenAI firms. Team from Tsinghua, ZJU, and returnees — unicorn founding experience and in-house algorithms.
Q&A
1. Product and core scenarios?
PromeAI.com — turn creative sketches into photoreal images and high-quality video for architecture, interior, product, games. Controllable generation + artistic expression: wild ideas with precise fidelity; everyday creators get master-level expression. Interiors: seconds-level style swaps (cream, Mediterranean, modern minimal, new Chinese) + walkthrough video. E-commerce: background swap, hard-to-cast models (lingerie, Black, plus-size), quick expand-to-platform sizes.
2. R&D / GTM; biggest 2023 challenge and win?
Steady R&D/GTM. Challenges: broad designer acceptance (tight designer/partner feedback loops) and brand share in a crowded market (shows, design communities, marketing). Harvest: many designers/creators, strong word-of-mouth, partnerships with top domestic/overseas studios.
3. GenAI opportunities vs prior waves?
More innovative, intelligent products; automate creative processes; LLMs as a “brain” calling other AI as hands/feet. Vs past: tech more available; almost every app can be redone; interaction upgrades mouse → touch → natural language.
4. China landing challenges?
Data/algorithm availability (esp. sensitive data); fast tech churn; law/ethics/privacy; intense competition requiring differentiation.
5. Key product / GTM factors?
PMF first — which users, which problems; what GenAI is/isn’t good at; balance tech ceiling vs need; 2C vs 2B, entertainment vs productivity — then match channels.
6. Competing with giants?
Flexibility + specialization: agile new scenes; vertical expertise, data, experience; custom solutions; accumulate users.
7. Are LLMs still too weak?
Apps needn’t depend only on LLMs — combine other AI. Choose tools by scene.
8. Human–machine collaboration?
Jarvis-like imagination: natural interaction, mass design options, iterate on feedback. For enterprise automation today is awkward — flexible flows aren’t smart enough; fixed flows don’t need smarts. Watch the usability boundary, then go all-in.
9. 2024 tech to watch?
Large multimodal models (richer AI senses of the physical world); embodied intelligence + RL — closer to AGI.
1 CN image; attach on ingest. Signup CTA in CN omitted.
— END —