How Many AI Avatars Can One GPU Handle? Real-World Test Reveals 4 Avatars at ¥7,600 Each per Month
📝 Originally published (in Japanese) at forge.workstyle.tech . Building an Unmanned System for 3D Avatar Live Streaming We're developing an unmanned system where 3D avatars automatically handle live streaming. The system boots up a cloud GPU pod at the scheduled start time, the renderer assembles and streams the video, and then the pod is discarded when the segment ends. Since there's no human…
A recent experiment revealed that a single GPU can support four 3D avatars engaged in simultaneous live streaming, with each avatar costing approximately ¥7,600 per month to operate. However, the experiment also uncovered two critical factors that impact the success and feasibility of such a system: reliability and startup speed.
The first challenge was ensuring the system could run multiple avatars without interruption. Despite utilizing the same image for all avatars, the experiment discovered that some hosts crashed every 60 seconds. To address this issue, a mechanism was implemented to automatically switch to a different host when one failed, thereby maintaining uninterrupted streaming.
The second challenge involved reducing the time it took for the system to become operational. Initially, it took four minutes for the system to start streaming after initialization. To improve this, the system's startup speed was optimized, reducing the time from pod startup to streaming start from four minutes to a remarkably swift 95 seconds. This enhancement proved crucial in enabling the system to effectively switch between hosts when necessary.
The combined effect of these improvements was a more reliable and efficient system, capable of delivering high-quality live streaming services with multiple avatars. By addressing the bottlenecks in capacity, reliability, and startup speed, the experiment demonstrated that it is indeed possible to build and operate an unmanned system for 3D avatar live streaming, while also providing insights into the costs involved.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
