To keep live streaming experiences smooth and engaging, 17LIVE now uses Gemini to analyze and summarize user feedback in real time, pinpointing the most critical technical issues as they happen. Previously, vague feedback from users meant engineers had to sift through countless potential causes for their dissatisfaction, slowing resolution and impacting the viewer experience. With targeted observations and suggestions provided by Gemini, the 17LIVE team can now troubleshoot faster and more accurately – leading to a 30% increase in user satisfaction.

Gemini also analyzes logs and historical data stored in BigQuery to detect technical incidents before they escalate. When an issue is identified, a Gemini-powered AI agent automatically initiates the necessary actions and assigns tasks to the appropriate team members. This improved coordination allows 17LIVE's engineers to prevent problems proactively, requiring 30% fewer site reliability engineering resources.

As a result, 17LIVE now can better focus on creating new user experiences and adjusting them to meet user demands. When a new feature is released, the company uses Gemini to collect and summarize user-related data and evaluate how the feature performs and what can be improved. Before the adoption of Gemini, the 17LIVE team needed to wait until the end of a month to receive new features' performance reports. Now, with daily reports generated by Gemini, the team can quickly adapt its feature design to users' preferences and achieve better reception.

For example, to enhance interaction between live streamers and viewers, 17LIVE has built AI cohosts with Gemini to greet and chat with viewers individually on behalf of live streamers. During streaming sessions, AI cohosts can suggest or create topics based on viewers' profiles to enhance engagement. When inappropriate content appears, AI cohosts also act as moderators to take necessary actions.

"With Gemini, we're able to optimize live streaming experiences by resolving issues faster and creating an engaging, positive vibe in our streaming rooms," says Hsu. "The changes are immediately noticeable."

Since 17LIVE doesn't record live streaming sessions, it previously had limited visibility into what content resonates most with users. To better understand viewer interests, the company now uses Gemini and Vertex AI to extract insights directly from live sessions. These insights help to generate detailed personas for both viewers and streamers, enabling more accurate content matching and personalized recommendation. This way, users are discovering content they enjoy more easily, resulting in a 10% increase in comment rates and a 20% higher follow rate.

On top of that, 17LIVE applies this persona data to generative search engine optimization, so that users who have not signed up for 17LIVE can quickly find the live streaming content that might interest them through generative AI-powered search. 17LIVE's live streamers are, therefore, able to expand their audience reach, realizing 10% higher external exposure.

17LIVE also uses Google Cloud video-generation technology to enhance the live-streaming experience. Its virtual streamers – avatar-based performers known as "VTubers" – appear on screen as animated characters. During live sessions, 17LIVE uses Gemini to analyze the livestream in real time and identify the emotions being expressed. That emotional data is then sent to Veo, which generates corresponding facial expressions for the avatars. Because these reactions update instantly and match the mood of the moment, V-Livers can create more engaging content and interact more naturally with viewers.

Next, 17LIVE will continue expanding its use scenarios of Google AI to create more engaging content, such as generating personalized visual gifts with Veo. It also plans to use Model Armor and other Google Cloud tools to safeguard generative AI security.

"The diverse AI capabilities of Google Cloud have transformed our operations and services, enabling glitchless, engaging, and interactive streaming," notes Hsu. "We're excited to explore more use scenarios and continue bringing surprises to our users."