The Ultra-Realistic AI Face Swapping Platform Driving Romance Scams
Summary
Haotian, a Chinese-language AI app that produces near‑perfect live face swaps and voice cloning, has been widely used in romance and cryptocurrency “pig butchering” scams. Marketed largely via Telegram, the service offered granular face and voice controls and streamed its output into popular messaging apps. Analysis by researchers and cryptocurrency tracers links at least $3.9m of USDT payments to Haotian, including funds tied to scam marketplaces and alleged criminal activity. After WIRED’s inquiry some Haotian channels disappeared.
Key Points
- Haotian produces high-quality, real-time face swaps and voice cloning that can be streamed into WhatsApp, WeChat, Telegram, Zoom and other platforms.
- The service was primarily marketed through a public Telegram channel and desktop subscriptions; promotional material emphasised social‑engineering use cases.
- Elliptic traced at least $3.9m in payments to wallets linked to Haotian; many payments had ties to sanctioned scam marketplaces such as Huione.
- Security researchers and UN reports identify Haotian as part of a broader scam tech ecosystem in Southeast Asia that supplies tools, stolen data and services to fraud networks.
- Haotian denies facilitating illegal use and claims entertainment use only, but evidence of marketing language and transactional links to scam services undermines that defence.
- Traditional deepfake checks (wave hands, look for glitches) may be less reliable as the tool claims to handle gestures, blinks and facial contact.
Content Summary
Haotian emerged around 2021 and offered users fine‑grained controls (up to 50 settings) to shape face swaps, allowing scammers to craft convincing personas for live video calls. The company sold subscriptions and advertised on Telegram, where it built a substantial following and provided updates and support. Researchers found direct links between Haotian and gray‑market services: payments flowed through cryptocurrency wallets, some of which interacted with Huione Guarantee — a Telegram-based escrow and marketplace later sanctioned for facilitating scam activity.
Independent analysis from firms such as Elliptic, Tehtris, and UN investigators documents Haotian’s place in a broader fraud tech stack used in Southeast Asia’s large scam economy. Although Haotian asserts it restricts illegal uses and claims to target entertainers and streamers, marketing materials and channel posts referenced social engineering and “deep chat” tactics used in pig‑butchering scams. Following WIRED’s outreach, key public Telegram channels and accounts associated with Haotian became inaccessible.
Context and Relevance
This story sits at the intersection of generative AI, platform moderation and organised online fraud. As deepfake tools get easier to use and cheaper to buy, scammers can scale highly believable social‑engineering attacks that culminate in large financial losses — especially in crypto. The Haotian case shows how small‑value tech purchases feed a much larger illicit economy and how messaging platforms like Telegram can be used to distribute and monetise harmful tools.
For readers: if you use dating apps or transact with new contacts online, expect fraudsters to increasingly rely on live deepfakes. Practical steps include insisting on shared, time‑stamped live verification (e.g. show a specific gesture with a piece of paper and the current date) and treating unsolicited investment or romance approaches with extreme caution. Regulators, platforms and investigators will need co‑ordinated responses to disrupt suppliers as well as front‑line scammers.
Author style
Punchy: This piece isn’t just another tech scare — it exposes a working business model that makes convincing fakes cheap and distributable. Read the detail if you want to understand how small transactions and Telegram channels prop up a global fraud machine.
Why should I read this?
Look, you don’t need to become a cyber sleuth — but you should know what the scammers are using. If you swipe, click or video‑call strangers online, this explains the exact tech that can make a liar look real. It’s short, sharp and saves you having to dig through dense reports yourself.