The Chinese company Tencent Cloud has unveiled a new digital platform for creating deepfakes, called people-Deepfakes-A-A-Service (DFAAS). According to reports from Chinese media, the service only requires a three-minute live video and 100 pronounced offers to create a high-clarity digital person for $145. The service takes only 24 hours to complete.
The digital person generated by the service can be tuned for some aspects such as skin tone and rear background. However, the generated deepfakes have a flat intonation and a single speech rhythm characteristic of traditional acoustic models. This is made possible through Tencent Cloud’s own technology for setting Temra with small samples based on acoustic models of deep learning and vocoders of neural networks.
Tencent Cloud plans to build an “AI + Intelligent Factory” for creating digital people and relies on a universal platform for production, sales, and maintenance. The factory for creating digital people is based on Tencent Cloud Ti, a machine learning platform that offers over 10 AI algorithms. The company offers five styles for creating digital people: 3D-realistic, 3D-semi-multfilm, a 2D-real person, and a 2D-multfilm. It’s also possible to create individual questions and answers for a digital person, turning it into a kind of chatbot.
Tencent is primarily focused on commercial applications for this technology in China. Local media reports suggest that Tencent could create digital doctors, lawyers, and other specialists with this technology.
| Generated Deepfake | Real Person |
|---|---|
 |  |