— By the end of 2025, China Telecommunication Artificial Intelligence Institute (TeleAI)announced a major breakthrough. Generated Video Compression (GVC). This technology leverages generative AI’s visual priors to extract and transmit compact perceptual features and motion dynamics. On the receiving end, it leverages the powerful generation capabilities of large-scale models to “infer” and reconstruct high-fidelity video with realistic textures and recovered details in real time. GVC achieves an unprecedented compression ratio as low as 0.02% while maintaining visual clarity. Extraordinary temporal smoothness. it makes possible Seamless, no delays Enable video calling, live streaming, and task-oriented video understanding even in weak signal or bandwidth-constrained environments, such as remote locations or offshore locations.
Classic communication theory, rooted in the Shannon-Weaver model introduced in the 1940s, conceptualizes communication across three levels. Level A deals with technical data transmission. Level B considers semantic meaning. Level C focuses on efficiency or task-oriented communication. For decades, video communications has primarily focused on optimizing Level A or B to maintain pixel fidelity. However, traditional methods often fail on weak networks, resulting in buffering, freezing, or pixelation when there is insufficient bandwidth. GVC closes this gap by shifting the focus to Level C and prioritizing the effectiveness of the information provided.
“In the era of AI, the nature of communication has evolved from data transmission to intelligence delivery,” said Professor Xuelong Li, CTO and Principal Scientist at China Telecom and Director of TeleAI. “GVC heralds a new phase in video communications, moving the core logic from ‘pixel transfer’ to ‘semantic generation’. By trading computation for compression ratio, it opens a viable path to advanced video communications paradigms that are effective, efficient, scalable and practical.”
build on AI flow TeleAI’s research team introduced the framework developed by Professor Xuelong Li in 2024. Law of information capacity Measure the knowledge density and intelligence capabilities of your model through data compression. At the heart of GVC is the core philosophy of trading compute for bandwidth. GVC leverages model distillation, quantization, and strong pregeneration to allow the receiver to synthesize missing details rather than transmitting them. This significantly reduces communication overhead and enables low-latency inference even on consumer-grade GPUs. As a result, the video stream remains visually smooth and consistent, completely eliminating the “stuttering” and “frame skipping” typical of low-bandwidth connections.
Unlike traditional codecs like HEVC, which struggle to preserve every pixel, GVCPerception + Task Orientation” mode sends only “perceptual guidance” capabilities for human vision and “semantically effective” representations for machine understanding. In standard benchmark tests (MCL-JCV), GVC: 0.005 bpp (~0.02%) While achieving a visually appealing reconstruction. Even under such extreme compression, the generated video maintains smooth motion and consistent semantics, ensuring a high-quality experience free of artifacts typically seen with low-bitrate streams.
In conclusion, GVCs China Telecommunication Artificial Intelligence Institute (TeleAI)significantly reduces transmission bandwidth requirements while maintaining high perceived quality. This is particularly suitable for scenarios where high efficiency and efficiency are required. Uninterrupted liquiditymaritime communications via narrowband satellites, emergency rescue operations, remote video surveillance under narrowband mobile networks, edge computing in vehicles or wearable devices, etc. By transmitting only the core information needed for perception and decision-making, GVC not only defines a new technology paradigm, but also redefines the nature of communication by ensuring smooth, intelligent, and adaptive connectivity regardless of network conditions.
For more information, China Telecommunication Artificial Intelligence Institute (TeleAI)click here: https://www.teleai.com.cn/product/AboutTeleAI
To read the technical report of Generated Video Compression (GVC), click here:
https://www.arxiv.org/abs/2512.24300
Contact information:
Name: Tang Siya
Email: Send email
Organization: TeleAI
Website: https://www.teleai.com.cn/product/AboutTeleAI
Release ID: 89182164
If you find any problems, problems, or errors in the content of this press release, please contact us. [email protected] to notify us (it is important to note that this email is the authorized channel for such matters; sending multiple emails to multiple addresses does not necessarily help expedite your request); We will respond and rectify the situation within the next 8 hours.
