Uploaded April 2026 | Updated September 2026, 2 weeks ago
Add a Gemini 3 AI participant to a video conference with Fishjam.
Adrian from Software Mansion walks through how to connect the Gemini 3.1 Flash Live audio model as a real participant in a Fishjam video conference. It can hear, see the camera feed, and talk back in real time, with Google Search, multilingual switching, and guardrails all built in.
What's covered: Setting up Fishjam and Gemini Live API, creating rooms and peer tokens, building a React frontend with the Fishjam SDK, connecting an agent to the conference, streaming audio and video frames to Gemini, handling interruptions, enabling Google Search, and using system prompts to define and restrict agent behavior.
You'll need a Fishjam sandbox account and a Gemini API key to get started. Full example code linked below.
Resources:
Live API docs → https://goo.gle/4dtm7eF
Fishjam x Gemini docs → https://goo.gle/47LorKk
Full example from the video →https://goo.gle/4caHfDY
Speaker: Adrian
Products Mentioned: Google AI, Gemini
Add a Gemini 3 AI participant to a video conference with Fishjam.
Adrian from Software Mansion walks through how to connect the Gemini 3.1 Flash Live audio model as a real participant in a Fishjam video conference. It can hear, see the camera feed, and talk back in real time, with Google Search, multilingual switching, and guardrails all built in.
What's covered: Setting up Fishjam and Gemini Live API, creating rooms and peer tokens, building a React frontend with the Fishjam SDK, connecting an agent to the conference, streaming audio and video frames to Gemini, handling interruptions, enabling Google Search, and using system prompts to define and restrict agent behavior.
You'll need a Fishjam sandbox account and a Gemini API key to get started. Full example code linked below.
Resources:
Live API docs → https://goo.gle/4dtm7eF
Fishjam x Gemini docs → https://goo.gle/47LorKk
Full example from the video →https://goo.gle/4caHfDY
Speaker: Adrian
Products Mentioned: Google AI, Gemini
