Tacheng AIGC Newsroom
Tacheng Media Convergence Center
Domain-model AIGC, from topic to broadcast.
An intelligent editing system built on a self-developed domain model — bringing AI topic selection, writing, and one-click production to the newsroom, cross-modal asset retrieval to the archive, and virtual digital-human anchors to broadcast.
Overview
The Tacheng Region AIGC project builds an intelligent editing system on the center's existing production and media-asset platforms, using Sobey's self-developed Mingmou domain model. It brings AI topic selection, writing, and one-click production to the newsroom, cross-modal retrieval to the archive, and virtual digital-human anchors to broadcast.
Challenge
A border-region media center needed to raise content production efficiency and accuracy while serving multilingual audiences — and modernize broadcast without losing the familiarity of its real hosts.
Faster, accurate production
Topic selection, writing, and production had to accelerate while keeping content timely and accurate.
Multilingual editing
Manuscript editing had to serve an ethnically diverse region across multiple languages.
Flexible broadcasting
Broadcast needed a more flexible, vivid presentation without losing the identity of the station's hosts.
Solution
Sobey's self-developed domain model drives the news process end to end — topic selection and planning, manuscript editing, video editing, news dubbing, and video broadcasting. Cross-modal retrieval brings semantic search to the media-asset layer, and a virtual digital-human application replicates the station's real hosts, letting virtual anchors broadcast like people with adjustable expressions, movements, and voice. Manuscript editing supports Chinese, Uyghur, and Kyrgyz for the region's multilingual needs.
Delivered with
The Sobey solutions and products behind this deployment — each built to work together as one chain.
Result
Tacheng's newsroom produces faster and more accurately with domain-model AIGC — multilingual editing, semantic asset retrieval, and lifelike virtual anchors from topic to broadcast.
News stages automated by the domain model — topic, writing, editing, dubbing, broadcast.
Languages in manuscript editing — Chinese, Uyghur, and Kyrgyz.
Virtual digital-human anchor replicating the station's real hosts.