skills.rest
amoai-tech
amoai-techCOMMUNITY
Explore all skills by this author →
GitHub

gemini-live-api-dev

Build real-time bidirectional audio and video streaming applications with the Gemini Live API over WebSockets.

1|Updated Sep 15, 2026
Search Tags:#websockets#speech-to-text#real-time audio#google-genai#gemini live api#voice streaming#live translation
One-click install
npx skills add https://github.com/amoai-tech/mdeai --skill gemini-live-api-dev-amoai-tech
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: gemini-live-api-dev
Source: https://github.com/amoai-tech/mdeai/tree/main/.claude/skills/gemini/references/official/gemini-live-api-dev
Command: npx skills add https://github.com/amoai-tech/mdeai --skill gemini-live-api-dev-amoai-tech

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires google-genai, @google/genai.

What problem does it solve? Building low-latency voice and video assistants requires managing WebSocket sessions, audio formats, interruptions, and authentication correctly. This Skill provides the operational knowledge to implement real-time streaming with the Gemini Live API without trial-and-error against the protocol. ## Core Features & Use Cases - Real-time streaming setup: Connect to the Live API with Python (google-genai) or JavaScript (@google/genai), send audio, video, and text via send_realtime_input, and process multi-part server events including transcriptions and interruptions. - Specialized live modes: Configure live speech-to-text transcription with gemini-3.5-transcribe-live and real-time translation across 70+ languages with gemini-3.5-live-translate-preview. - Production concerns: Handle session resumption, context window compression, ephemeral tokens for client-side auth, and migration from deprecated Gemini 2.5 Live models. - Use Case: You are building a voice concierge in a Next.js app. Use this Skill to connect a browser client with ephemeral tokens, stream microphone PCM audio at 16kHz, receive 24kHz audio responses, and clear playback queues on interruption signals. ## Quick Start Ask the AI to create a Gemini Live API session in JavaScript that streams microphone audio and plays back spoken responses with interruption handling.

Frequently Asked Questions about gemini-live-api-dev

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I connect to the Gemini Live API in JavaScript?▼

Use the @google/genai SDK and call ai.live.connect with a model like gemini-3.1-flash-live-preview, response modalities, and callbacks for onmessage, onerror, and onclose. Send input with session.sendRealtimeInput using audio, video, or text keys.

How do I stream microphone audio to the Gemini Live API?▼

Capture microphone input as raw PCM, little-endian, 16-bit, mono at 16kHz, then send chunks via send_realtime_input with MIME type audio/pcm;rate=16000. Output audio arrives as 24kHz PCM that you queue for playback.

Does the Gemini Live API support WebRTC?▼

The Live API only supports WebSockets natively. For WebRTC, use partner integrations such as LiveKit, Pipecat by Daily, Fishjam, Vision Agents by Stream, Voximplant, or the Firebase AI SDK.

Why is my Gemini Live session disconnecting after 10 minutes?▼

Live API connections have a roughly 10-minute lifetime and audio-only sessions last 15 minutes without compression. Implement session resumption and enable context window compression to handle resets and extend conversations.

Can I use both text and audio responses in one Live API session?▼

No, each session supports only one response modality, either TEXT or AUDIO, not both. Native audio models like gemini-3.1-flash-live-preview only support audio output, though you can enable output audio transcription for text transcripts.

How do I migrate from Gemini 2.5 Flash Live to Gemini 3.1?▼

Update the model string to gemini-3.1-flash-live-preview, replace thinkingBudget with thinkingLevel, process all parts in each server event, and use send_realtime_input for conversation text. Remove proactive audio and affective dialogue configurations, which are not yet supported.

Related Skills

Explore More Software Engineering Skills→
heyu-aiheyu-ai

verify-gemini-models

Verify Gemini model availability and functionality via REST and WebSocket calls.

Official
Advanced
Lucmac667Lucmac667

ASR

Transcribe spoken audio into text using the z-ai-web-dev-sdk.

Community
Intermediate
spideynolovespideynolove
2

embedding-strategies

Compare embedding models and define chunking strategies for RAG pipelines.

Community
Advanced
jokken79jokken79
1

voice-ai-development

Develop real-time voice AI applications using OpenAI Realtime API, Deepgram, ElevenLabs, and LiveKit.

Community
Advanced