---
title: "Unique Interview | Niren Intelligence CEO: Third-Generation AI + Vision Human–Computer Interaction"
author: "Unique Research"
sourcePublication: "Unique Research (FFCAP English library)"
originalPublishedAt: "2024-09-11T16:00:00+08:00"
canonical: "https://ffcap.cn/en/research/20240911a05iar00"
source: "https://ffcap.cn/en/research/20240911a05iar00"
language: "en"
cnSourceUrl: "https://view.inews.qq.com/a/20240911A05IAR00"
---

# Unique Interview | Niren Intelligence CEO: Third-Generation AI + Vision Human–Computer Interaction

_Unique Research · Interview · September 11, 2024 · Interviewee: Dr. Liu Gang, CEO, Niren Intelligence (拟仁智能) · Interviewer / editor: Qiuping_

In China’s crowded, homogenized digital-human lane, one company kept building through the metaverse boom-and-bust: Niren Intelligence. Founded end-2020, it combines AI and computer graphics for high-fidelity, personified, personalized AI virtual humans. Dr. Liu’s line in the sand: only a 3D virtual human that supports **both AI drive and real-time human takeover** counts as a “metaverse virtual human” — able to live in 3D virtual, real, and metaverse scenes with natural interaction.

## Products and deployments

**Q: You say only 3D virtual humans with both AI real-time drive and human takeover qualify as metaverse virtual humans — still hard technically. Are you building that, and where are you?**

**Liu:** We are on that path. In basic scenarios — livestream and customer service — we already hit the target. Livestream is exploring “auto livestream” 24/7 via AI, but some segments still need a human (voice and/or on-camera drive of the front-end). Our stack supports that. In CS, the default is unattended bots; when a user is emotional or rejects the bot, we hand off seamlessly to a human agent. So dual AI-drive + human-takeover exists in both.

Richer 3D virtual spaces are harder: full-body mapping (fingers, joints, micro-expressions, gaze) for human drive; for AI drive, environmental perception → analysis → decision → feedback — unlike livestream/CS. Long road; we start small and iterate toward the full vision.

**Q: Which industries are live today?**

**Liu:** Core focus is intelligent assistants. Explorations in smart manufacturing — e.g. EVs putting a highly interactive virtual assistant on the center screen (windows, HVAC, etc.). Smart home / kitchen: ovens, fridges, range hoods with touch screens — a small hood screen as a cooking/nutrition expert you wake and talk to. These are demand-driven products, not pure imagination.

**Q: What can your digital-human SaaS do?**

**Liu:** Built for enterprises and government. Two image asks we always hear: custom brand avatars, and online outfit changes (festivals, etc.). The platform ships festival/traditional costumes and props, plus user-uploaded assets published live without re-shipping the app. Knowledge-base updates are live too — change company registration copy in the admin and the next question about business scope gets the new answer immediately.

**Q: Do industries differ much on underlying AI / LLM needs?**

**Liu:** Yes. Outsiders call us a digital-human company; we call ourselves an AI company — customized intelligence per industry/enterprise, not just avatars. We finetune on a general foundation model per customer; we do not build the “layer-zero” base model (capital-heavy). We own the middle and application layers that turn AI into industry fit.

## Pain points

**Q: Crowded market, but UX breakthroughs feel rare — why?**

**Liu:** Many chase expressiveness and conversational fluency — hair strands, lively eyes. Critical for games/film; elsewhere the AI interaction and real service capacity behind the face matter more. Value is whether the knowledge and skills solve real problems. Industrial pipelines can make pretty avatars; interactive, real-time problem-solving is still thin. We started before the metaverse hype; the concept cooled and drew skepticism; we still talk metaverse because it is the destination.

**Q: Thin personalities and weak problem-solving — your view?**

**Liu:** Stand on solving real problems first; then pursue higher expressiveness. Distinct personality is worth long R&D, but industry priority today should be utility over perfect performance.

**Q: Avatars that don’t add brand value — how to fix?**

**Liu:** Brand personality travels better through a character than through copy alone. Haier Brothers spent heavily on animation not just for ROI but to lodge an image users map to the core business — and they keep personifying it. That is a long build; we are in that process with the industry.

**Q: Missing a Vision Pro–class platform to host digital humans?**

**Liu:** Every new medium needs maturity. 3D characters × AR/VR glasses is huge — e.g. light AR glasses on a tourist, vision/GPS/BeiDou place recognition, a digital guide narrating history as you walk. Digital infra and digital-human tech are getting ready; hardware — cost, thinness, comfort, adoption — is the gate. Not far, as hardware improves.

**Q: Still mostly single-scene, not full process automation or full real-world replication?**

**Liu:** Don’t dismiss any scene — film/TV efficiency and cycle cuts are real wins. For Niren, the priority is real-time user interaction that solves the problem in front of the user, personalized and immediate.

**Q: Delivery challenges across unique customer needs?**

**Liu:** Unlike a VCR promo (feedback is “nice video”), our work is judged on whether it fixes business problems and how smart it feels. LLMs and AIGC are central. We go deep with large enterprise customers to learn real needs — not project-churn for cash-flow — then generalize into a platform from those scenes. Personalized asks → broader solutions → platform company.

**Q: Codifying best practice then replicating — like consulting, heavy know-how?**

**Liu:** Extremely hard. Depth in one industry with stable data and solution muscle is already an achievement. Don’t expect a do-everything AGI soon; for years, what actually fixes business will be industry- and domain-specialized. Better: enterprise models, then personal models — not only industry models. Giants with compute can own layer-zero / industry foundations; enterprise- and person-level personalization is still under-served — early depth there is the opportunity.

**Q: Biggest challenge for you and the company?**

**Liu:** The gap between customer expectations of AI and what delivery can do. Users want AI that outthinks humans on every problem; models often sound plausible but lack precision at org/person level. First align: there is no perfect AI that solves everything. Then the engagement is joint — win-win or lose-lose, not vendor-only heroics. That partnership mindset is how progress happens.

## Trends

**Q: Domestic vs overseas digital-human trajectory?**

**Liu:** Many valuable scenes. We focus on AI-driven real-time interactive virtual humans; human-takeover is a separate discussion. Dig into concrete business needs so the tech is not a gimmick.

**Q: Industry still lacks an ideal product for that need?**

**Liu:** We keep shipping tools. Paradox: every enterprise wants digital human + LLM to solve all business problems, while business data changes every day — no third-party vendor can live inside forever retraining. So “teach to fish”: give enterprises usable tools and methods to empower themselves.

**Q: What drives you?**

**Liu:** Second startup after years in multinationals. Fuel is building with Gen-Z / Gen-Alpha teammates in a lane that will mint one or more great companies. From day one the goal was **third-generation HCI**: not gen-1 keyboard/mouse text, not gen-2 voice — but vision-based interaction. Face-to-face with a digital human is a different experience from voice commands. Motivation is being inside something that may set the next default.

**Q: Consumer hardware partners for something popular and fun?**

**Liu:** Some PC/phone talks are already live. Consensus is growing that next-gen interaction is not only swipe.

**Q: Phones add AI features but still feel limited?**

**Liu:** For a true third-gen system, the architecture must be fully open — voice (and vision) able to invoke any underlying app, none excluded. Revolutionary; not overnight.

**Q: Hardware today is still a broadcast medium — how does it become an agent that “speaks”?**

**Liu:** Vision: a phone that is a daily interactive partner, learning your patterns until it knows you better than most people. Abstract that into a embodied agent and it approaches your digital twin — not mere data archival, but an agent of behaviors, habits, even thinking. That is the “digital immortality” we mean — wider life and mental space for every individual. I hope we get there.

— END —

---

Editorial English edition published on UniqueCapital: https://ffcap.cn/en/research/20240911a05iar00
On-site reading page: https://ffcap.cn/en/research/20240911a05iar00