On February 25th, JD Retail's technology team announced the official launch and full opening of its self-developed Jing Dian Dian AIGC content generation platform. This platform aims to provide e-commerce merchants with an efficient and cost-effective content generation solution using AI technology. Currently, it covers over 20 core scenarios, with daily AI capability calls exceeding 10 million. It assists over 350,000 JD merchants in generating high-quality product images, marketing copy, and upcoming main video images with a single click, significantly improving content production efficiency and reducing costs.
The launch of the Jing Dian Dian AIGC platform marks a significant technological breakthrough for JD in the field of e-commerce content generation. Based on various AI capabilities, this platform revolutionizes traditional e-commerce content production methods, covering images, copy, and videos. Without requiring professional expertise, ordinary users can easily generate professional e-commerce content materials. Currently, Jing Dian Dian has launched two core AIGC capabilities: AI product image generation and AI marketing copy generation.
For AI product image generation, users only need to upload ordinary product photos or white background images. The system automatically performs background removal and, using e-commerce data, recommends suitable scene templates to generate high-quality product scene images. Additionally, the system can add core selling points and marketing benefits, generating product main images, detail images, and marketing images. AI marketing copy generation allows users to input JD product SKU numbers or names. The system extracts selling points from related products and generates precise marketing copy based on the user's desired writing style.
To overcome technical challenges in AI content generation for e-commerce, Jing Dian Dian has implemented several technological innovations. First, the platform trained JD's text-to-image base model using massive retail image data, employing the DiT (Diffusion Transformer) framework and Flow Matching technology to significantly improve image generation efficiency and quality. Second, Jing Dian Dian independently developed ReferenceNet and ControlNet to achieve precise injection of image feature consistency and precise control over image contours, style, and layout, ensuring the realism and consistency of generated images. For marketing copy generation, Jing Dian Dian has developed a self-owned multi-modal product understanding model, combining the RAG (Retrieval Augmented Generation) approach and product knowledge to generate factually accurate and stylistically diverse marketing copy. Furthermore, the platform incorporates a reinforcement learning mechanism to optimize the generation model based on user feedback and product data, improving content production quality.
The Jing Dian Dian AIGC platform has demonstrated significant effectiveness. Currently, it's fully open to JD merchants, ecosystem partners, and internal employees, integrated into JD's core B-end products, such as intelligent background removal, intelligent copywriting, and product scene image generation. Users have experienced over a 95% increase in efficiency and over a 99% reduction in cost for product image and marketing copy creation. For example, creating a 2D home decoration scene image traditionally required a week and over 10,000 yuan. Jing Dian Dian AI image generation only requires uploading a few product images to quickly generate high-quality images without needing professional designers or physical space.
Jing Dian Dian also innovatively uses a large model acceleration scheme and a combination of large and small model inference technology to significantly improve resource utilization efficiency. Compared to traditional single large model solutions, Jing Dian Dian reduces resource input by up to 90% while maintaining consistent content generation quality. Through modular design and domain knowledge injection mechanisms, Jing Dian Dian ensures high adaptability and scalability across various e-commerce scenarios.
Website Experience: https://ai.jd.com
