Unique Research · Interview · September 25, 2024 · Guest: Li Shaobo, Founder & CEO, Tianxin Technology (甜新科技) · Interviewer: Wei Wu, Partner, Unique Capital · Edited by Qiuping
Since 2016, Tianxin has built large-model solutions at the intersection of AI / GenAI and 5G video communications — scenario services that fuse AI with video for brand clients. Unique Research spoke with Li about AI video in marketing and why 5G + AI video outbound robots beat classic voice dialers.
Q: Why video marketing for your second startup?
A: Two filters: understand the industry’s business model; stay in domains you truly know. From telecom plus 2016–17 video growth with 4G, we applied A/V to finance — digitize face-to-face sales, CS, and compliance. Finance clients are sticky (healthy collections, high pay/renewal) once you’re in — high barrier, low churn.
The ceiling was industry standardization and growth. COVID + Douyin short video / livestream trained consumers on video marketing — high information density, efficiency, UX. Stay time on short-video apps now dwarfs classic e-commerce and social. With 2022 mass 5G, carrier networks began supporting videoized services; SMS/voice tilt toward video. China internet waves track infra upgrades; 2022’s twin upgrades — 5G build and compute — mark a burst opportunity. From 2022 we pivoted fully to marketing: the “5G video marketing” era — realtime A/V + AI for interaction.
Q: AI wave vs mobile-internet wave — bigger opportunity?
A: Roughly peer in scale of chance, but AI’s change will outstrip mobile. Mobile moved users PC → phone — convenience and frequency, not a new interaction paradigm (quantity). AI is a quality jump — intelligence reshaping production, service, decision-making. Impact varies by sector, but commercial upside is large.
Q: 5G realtime short-video forms — e.g. digital-human outbound? Demand?
A: Telemarketing is old and still carries huge demand — but low information density and low intelligence create “spam SMS / robocall” pain. Merchants still need reach; if density is low they raise frequency, which backfires. Elevator ads or Douyin ads rarely feel like harassment because of timing, scene, UX, and relevance. So raise density, UX, intelligence — plus relevance and benefits (celebrity endorsement, valued content) so a call can feel like a welcome service.
Q: When realtime video meets AI / LLMs?
A: People connect with smart, positive counterparts; pleasant brand interaction lifts engagement. Early cold bots killed talk-willingness. Short video and livestream prove emotion drives consumption when content/people fit. In CPG, emotional value is decisive — quality + fair price + right moment → higher conversion. New interaction modes + intelligence + efficiency will reshape the industry.
Q: Can you match GPT-4o realtime voice yet?
A: Not yet. In video marketing the hard problem is personalization (“thousand faces”) — strong predictive models before the interaction (voice, copy, benefits, even color). Realtime interest scoring and content gen/pre-gen are hard; we’re accumulating data. China buyers care more about practical results than bleeding-edge stories — a clear China vs overseas difference.
Q: Vs classic voice AI dialers?
A: Two logics: upgrade an existing market, or open a new one. China’s commercial SMS + smart AI outbound ~¥400B/year. Video marketing beats pure voice on effect; ROI decides adoption. We price at parity with voice outbound to cut decision friction — heavy cost work so the switch is easy; acceptance has been relatively high.
Q: Incremental markets beyond share gains?
A: (1) Verticals where voice AI outbound never worked well — especially consumer, where video outbound ROI beats finance because visual content amplifies value; Douyin-scale ad spend shows the pull. (2) Content marketing for series/films/short dramas — natural video fit. Short video in China is ~10 years in; traffic costs stay high and saturated — advertisers want cheaper, better-UX, smarter video inventory vs rising mega-platform CPMs.
Q: Beyond phone calls — other 5G realtime video landings?
A: Carrier channels (SMS→5G messaging with video/interactive dialogue — now including Apple devices) and internet channels (WebRTC / H5 realtime). Ads and services are shifting from text/static pages to visual interactive scenes.
Q: Will every ad, landing page, and site support realtime video?
A: Trend is hard to reverse. Modern browsers support mature WebRTC (~10+ years). GPT-4o uses WebRTC as core tech; streaming LLMs + video synth push the same way. Humans prefer vivid, intelligent, fast agents.
Q: Relation to multimodal model vendors?
A: Ecosystem leans on big model labs for core tech — streaming LLMs, multimodal, realtime video gen are capital-heavy. Scene customization stays with industry specialists; if labs build every niche app, economics fail. We stay focused: high A/V demand, strong interaction/UX, high AOV marketing/finance — not every industry needs complex interaction; text CS may suffice for info-display verticals. Balance capability vs cost and price to value.
Q: MarTech/SalesTech competitors; more product for user management?
A: Possible — we may expand user-management features. MarTech/SalesTech value = marketing-efficiency lift. SCRM once lifted efficiency, then habit shifts eroded it. Three efficiency levers: (1) video > text/voice (info density ~100× — 15s video ≫ 15s audio; livestream e-com vs graphic e-com); (2) AIGC lifts video production efficiency another order of magnitude; (3) personalization at the marketing entry. Competition then is execution: industry insight, team, product understanding.
Q: Unique competitiveness?
A: ToB enabling brands to serve C better — map B and C pain, clarify interest with cloud/LLM vendors (they want more cloud/token burn; realtime dialogue and video burn far more than typing). Org capability is as hard as product. Hiring the right specialists when you pivot matters most.
Q: Future of realtime video interaction?
A: (1) IM markets still huge — phone value is the screen; screen-share (“seeing beats hearing”) is the most-used meeting feature. (2) Video outbound likely becomes a communication default — screens have natural ad attributes. (3) Smarter voice agents will deepen phone-screen interaction value.
— END —