🤖 Wink and Rostelecom Hire AI Editor

The new digital assistant is designed to automatically populate incomplete profile cards for actors and content creators, a task particularly crucial for the archival segment of Wink’s library, which spans over 80,000 titles.

The agent is capable of independently gathering biographical data from disparate sources and performing “blind” restoration on legacy or low-quality images, converting them into reference-standard portraits. The deployment of this technology aims to increase search relevance, optimize the performance of recommendation algorithms, and reduce the time viewers spend choosing a movie.

Oleg Nikitin, Chief Technology and Product Officer at Wink, explained that the neural-network-driven editor operates entirely autonomously without human intervention. For image processing, it utilizes a computer vision algorithm that reconstructs faces pixel-by-pixel—even from blurry film frames—based on 68 control coordinates. This process removes artifacts, foreign objects, and sunglasses, while upscaling the resolution up to 4x. Concurrently, a proprietary Natural Language Processing (NLP) algorithm conducts multi-stage verification of textual data across several trusted sources to eliminate errors arising from identical names, subsequently formatting biographies to match the platform’s editorial standards. If the final automated quality control check issues a low score to a processed photo, the agent triggers a self-correction cycle and begins working on an alternative image.

The technological foundation of the AI editor consists of an ensemble of independent neural networks deployed within Rostelecom’s “Neuro-Gateway,” utilizing the GPU capacities of “Turbo Cloud.”

According to Natalia Musorina, Director of Client and AI Vendor Relations at Rostelecom’s IT Unit, the project is the result of synergy between Wink’s product expertise and the broadcaster’s experience in building complex AI architectures.

The new tool will operate in tandem with the “Entertainment Agent” launched in late 2025, helping it deliver more precise personalized recommendations and provide viewers with instantaneous, detailed information regarding production crews.