The Pika AI video call feature is here, and it changes how people interact with artificial intelligence agents. Pika has launched a beta skill that lets users send a Google Meet invite to any AI agent and hold a real-time, face-to-face video conversation with it. No more typing prompts into a chat box.
The feature is powered by PikaStream1.0, Pika’s new real-time streaming model. It works with any agent, including Claude, OpenClaw, and Hermes, making it one of the most open video chat tools in the AI space right now.
What the Pika AI Video Call Skill Actually Does
The skill is available as an open-source download on GitHub under the name pikastream-video-meeting. Once installed, the bot joins a Google Meet session as a real-time avatar. Users can provide their own avatar image or generate one using OpenAI image models. Voice cloning is also supported: record a short audio clip, and the agent speaks in a cloned version of your voice.
Memory and personality carry over from session to session. The agent does not reset between calls. It knows who you are, what you have been working on, and who else it has spoken with. That continuity is what separates this from a simple voice assistant.
Agentic Tasks During the Call
For users on Pika’s AI Self platform, the experience goes further. While you are talking, the agent can execute tasks in the background. It is not just listening and responding. It is working. Think of it as a productivity tool that holds a conversation and gets things done at the same time.
The bot also synthesizes workspace context into every conversation. It pulls in your identity, recent activity, and known contacts to make the dialogue feel natural and informed rather than generic. After the call ends, it automatically retrieves and shares meeting notes.
Built-In Billing and Setup
Pika built automatic billing directly into the skill. Before joining a meeting, the bot checks your account balance. If funds are low, it generates a payment link and waits for the transaction to complete before proceeding. It is a small detail, but it shows the team is thinking about real-world deployment, not just demos.
Setup involves three steps: generate or upload an avatar, optionally clone a voice, and send the agent a Google Meet invite. Pika AI Self users get the full agentic experience out of the box. Everyone else can grab the open-source skill from GitHub and connect it to their preferred agent.
Why This Matters for the Future of AI Agents
Text-based chat has been the default interface for AI agents since the beginning. Pika is betting that video changes the dynamic entirely. Conversations with a face and a voice tend to feel more natural, more engaging, and more productive. That intuition is backed by decades of research on human communication.
The Pika AI video call launch is a beta, so rough edges are expected. But the underlying idea, that your AI agent should be able to join a meeting just like a colleague, points toward a significant shift in how smart technology fits into daily work. Typing to an agent may soon feel as dated as sending a fax.
“Conversations tend to go better with a face and a voice. That’s why we’re thrilled to release the beta version of the first video chat skill for ANY agent.”
Pika Team
