In the current digital media era, video and audio have become important content trends. However, producing these forms of content is often more time-consuming than text, and existing text workers may lack the relevant skills to produce multiple forms of content. Generative AI is the key to solving this challenge, bringing revolutionary possibilities to media production.
In the process of trying to use generative AI, the Jusi team also applied for the nDX Taiwan News Digital Transformation (hereinafter referred to as nDX). This program was commissioned by Google from the Digital Economy and Industry Development Association to manage Google's "Taiwan News Digital Co-Prosperity Fund" and plans grant programs and implementation, with the goal of assisting Taiwanese news organizations to promote digital transformation and innovation. After Jusi received the grant, it will have more resources to explore and apply generative AI, improving the efficiency and diversity of content production.
Jusi Culture's AI collaboration attempt
With content as its core, Jusi Culture has conducted many meaningful experiments using generative AI:
- Basic technology for establishing multi-dimensional digital narratives: convert text into video and audio, and enhance the reading experience with multi-digit story delivery methods.
- Empowering editorial content production: not only defined in text, but also allowing editors to more easily convey content using images, charts and audio files.
- Automated application creation behind No Code: The digital story topics produced in the backend of No Code can deliver correct and valuable information to readers in more diverse forms of expression.
Jusi uses innovative functions of the editing platform to use multiple materials to tell stories
In this quiet revolution of large-scale newsrooms, we have also developed "virtual AI colleagues" to collaborate with humans on news production.
First, there are the AI virtual anchors "Wei Shidai" and "He Liwei": through the 5-minute daily report process, the news report of the day is launched. This process can meet readers on YouTube and Podcast platforms, allowing the media readership to add new listeners in addition to readers, breaking through the limitations of traditional text media.
The second colleague is a "virtual art designer": based on the information mentioned in the article, after understanding the meaning, he creates visual charts in just a few seconds to help readers understand the report content more easily.
The third virtual colleague is the "AI Digital Story Management System": it provides a cross-media conversion platform for reporters and editors to centrally manage audio, text, and image data for easy and quick recall.
"AI Digital Story Management System" utilizes a number of AI auxiliary tools, including:
- Image recognition: AI can recognize image text, convert and reformat it.
- Audio file transcription: Supports the rapid conversion of MP3 and other audio files into verbatim transcripts, and distinguishes different sound sources.
- Article generation: Quickly convert verbatim transcripts of podcasts into complete reports.
- Data visualization: AI identifies data, generates bar charts or line charts, and highlights key data changes.
- Illustration templates: Provides 6 templates, including community diagrams, comparison diagrams, mind maps, etc., to help readers better understand text content.
Study 1: How to make charts more in line with needs?
During the experiment, we learned that although AI can quickly generate text, the soul of the story, which key points should be amplified, which words should be marked, color matching and typesetting are currently done well by humans, and AI still relies on human editors. Unique insights and creative thinking.
In this creative ecosystem full of possibilities, the core value of human editing is: accurately capturing the essence of content, giving text soul and context, and controlling the key link of AI output quality.
Secondly, it is proposed that prompt is not only a technical operation, but also a dynamic process of seeking creative consensus. In this process, the wisdom and judgment of human editors are always the most valuable elements in AI-assisted creation.
Study 2: How to integrate AI charts and AI writing processes into colleagues' daily work?
In the process of introducing enterprise AI technology, we found that success does not lie in getting it right at once, but through progressive learning and adjustment, step by step to achieve everyone's expectations as much as possible.
Introducing AI tools into the editing room workflow can be roughly divided into three stages: "overcoming technical prototypes and establishing basic functions", "open beta stage", and "collecting feedback and optimizing prompts and UI/UX".
The key is to start with simple functionality, spread from a single department to multiple teams, and remain flexible and open-minded to let AI tools gradually integrate into the organization's work culture.
Learning 3: Stand firm on the shoulders of giants and be ready to run at any time
As technology giants compete, AI iterations become smarter, and many technical problems are gradually solved in new versions. This convenience makes media workers feel happy.
As the models become more powerful, the quality of the articles also improves. After each test, the performance of the new model always breaks through the limits of the old version, allowing us to deeply appreciate the possibilities brought by technological advancement and better meeting the editor's diverse narrative needs.
Results: The process was established, and after improving work efficiency, it was integrated into the daily editing workflow!
The nDX project has driven a quiet revolution in workflow at Jusi's editing desk. This one-stop solution seamlessly integrates audio files, text, articles and graphics, significantly lowering the technical threshold for content production. For editorial teams that have long been under time pressure, this means fundamentally changing the possibilities for content production.
Work that used to take a lot of time and effort can now be completed quickly through simple and intuitive operations. From converting audio files to text, processing text into articles, and supplementing it with data charts to increase readability, the whole process has become smoother than ever. This not only greatly stimulates editors' willingness to use charts, but also provides readers with a richer and more in-depth reading experience.
What's even more exciting is that the design of this project is highly flexible and inclusive. Whether it is front-end or back-end, colleagues from different departments can easily get started, and the scope of use is unprecedented. It is not only a tool, but also an empowerment - allowing editors to no longer be limited to pure text output, but to more naturally and easily integrate multi-element materials such as images, audio files, charts, etc. to completely reshape content. The imaginative boundaries of creation.
