---
title: "Gemini Live API Development: review, setup & limitations · Undominated.ai"
canonical: https://undominated.ai/skills/google-gemini-gemini-live-api-dev/
description: "Gemini Live API guidance for real-time audio, video and text over WebSockets, with stated session limits."
---

# Gemini Live API Development: review, setup & limitations · Undominated.ai

> Gemini Live API guidance for real-time audio, video and text over WebSockets, with stated session limits.

[← Explore all skills](/skills/)

AI APPLICATION DEVELOPMENT / Google

# Gemini Live API Development

Gemini Live API guidance for real-time audio, video and text over WebSockets, with stated session limits.

 [See setup guidance ↓](#setup)[Original source ↗](https://github.com/google-gemini/gemini-skills/tree/ee9da52195752ca83c2fca701d96bcb5a9c08d2d/skills/gemini-live-api-dev)

SOURCE REVIEW

 Reviewed 2026-10-13
 Evidence 5 linked sources
 Publisher Google
 Licence [Apache-2.0 ↗](https://github.com/google-gemini/gemini-skills/blob/ee9da52195752ca83c2fca701d96bcb5a9c08d2d/LICENSE)
 Revision ee9da5219575
 [Read what was—and wasn’t—checked ↓](#review)

“Use this skill when building real-time, bidirectional streaming applications with the Gemini Live API, or migrating legacy Live models (2.0/2.5/3.1) to Gemini 3.8 Live. Covers WebSocket-based audio/video/text streaming, voice activity detection (VAD), background reasoning (extended thinking), asynchronous function calling, session management, ephemeral tokens, live transcription, and live translation. SDKs covered - google-genai (Python), @google/genai (JavaScript/TypeScript).”

 [Google · upstream description ↗](https://github.com/google-gemini/gemini-skills/blob/ee9da52195752ca83c2fca701d96bcb5a9c08d2d/skills/gemini-live-api-dev/SKILL.md) Our analysis follows below.

01 / THE REASONING

## Why this made the selection.

 - Its Limitations section states session ceilings without compression (15 minutes audio-only, 2 minutes audio plus video, about 10 minutes per connection) and points to context compression and session resumption.
- Its best practices say to use ephemeral tokens for client-side deployments and never expose API keys in browsers.

### A good fit for

 - Building a voice or video agent on the Gemini Live API
- Migrating a legacy Live model to the Gemini 3.8 Live models the file names
- Handle interrupted audio and background tool completion without dropping other parts of a Live API event.

### Weigh up before choosing

 - The quick-start authentication examples pass a literal API key placeholder to the client, including in JavaScript, which the file's own best practices advise against for browsers. Use ephemeral tokens in client code.
- The file says MCP documentation results are always up to date and should be trusted as the source of truth. That claim was not checked.
- Session limits, model IDs and SDK floors (google-genai and @google/genai >= 2.3.0) are as written at this revision. The README states this is not an officially supported Google product.

02 / THE REVIEW RECORD

## What we actually inspected.

Source review has boundaries. A clear record is more useful than a “safe” badge.

### Material inspected

 - skills/gemini-live-api-dev/SKILL.md (models, SDKs, audio formats, quick start, limitations, best practices, documentation lookup)
- README.md (skill table, installation, disclaimer)
- .claude-plugin/plugin.json and .claude-plugin/marketplace.json (plugin source ./)
- LICENSE Apache-2.0 text
- skills/gemini-live-api-dev/SKILL.md
- README.md
- LICENSE

### Our findings

 - Input audio is raw 16-bit PCM at 16 kHz and output is 24 kHz, per the file.
- gemini-3.8-live-extended-thinking requires behavior NON_BLOCKING on every function declaration; blocking mode returns an error, per the file.
- The receive examples process all parts of a server event, clear playback on interruption and use interaction_status to distinguish ongoing background work from an idle extended-thinking session.

### Not established by this review

 - The installer was not run, and the skill was not executed against a project or service. No Live API session was opened.
- references/migration.md and the linked hosted documentation were not opened.
- Not executed: it was not installed or run.

The review applies to the material and revision named here. A newer upstream release can change its behavior.

03 / PUT IT TO WORK

## Add a skill to your workflow.

[Upstream setup instructions ↗](https://github.com/google-gemini/gemini-skills/blob/ee9da52195752ca83c2fca701d96bcb5a9c08d2d/README.md)

DOCUMENTED COMMAND

 /plugin marketplace add google-gemini/gemini-skills Copy command ↗

Copying does not execute this command. It may retrieve a newer version than the reviewed source.

 - The README has no skill-specific command for this skill. For Claude Code it documents /plugin marketplace add google-gemini/gemini-skills, then /plugin install gemini-skills@gemini-skills. The marketplace source is the repository root, so the plugin carries all three skills.
- Neither command pins revision ee9da52195752ca83c2fca701d96bcb5a9c08d2d. Compare the installed SKILL.md and references/migration.md with that revision.
- Do not ship a long-lived API key in browser or mobile code; the file directs client deployments to ephemeral tokens.

### Before you start

 - Claude Code for the plugin path
- A Gemini API key and google-genai or @google/genai only if you run the generated code

### Compatibility

Agent Skills-compatible hosts · Claude Code · Cursor · Codex · Antigravity · Gemini Live API over WebSockets

### Implementation

Python · JavaScript · TypeScript

### Host-mediated project files and Gemini Live sessions

 - Reads and writes application code in the working tree.
- Generated code opens WebSocket sessions to the Gemini Live API and streams the audio, video or screen frames the application sends.

### Cost model

Source is Apache-2.0. Live API usage is billed by Google to the user's account. Those charges were not measured.

04 / FOLLOW THE EVIDENCE

## The source trail.

Our notes are separate from the original resource. Check upstream before adopting a new version.

 - [Skill definition ↗](https://github.com/google-gemini/gemini-skills/blob/ee9da52195752ca83c2fca701d96bcb5a9c08d2d/skills/gemini-live-api-dev/SKILL.md) Checked 2026-10-13 https://github.com/google-gemini/gemini-skills/blob/ee9da52195752ca83c2fca701d96bcb5a9c08d2d/skills/gemini-live-api-dev/SKILL.md Supports: summary, upstreamDescription, whySelected, limitations, access, review
- [Plugin manifest ↗](https://github.com/google-gemini/gemini-skills/blob/ee9da52195752ca83c2fca701d96bcb5a9c08d2d/.claude-plugin/plugin.json) Checked 2026-10-13 https://github.com/google-gemini/gemini-skills/blob/ee9da52195752ca83c2fca701d96bcb5a9c08d2d/.claude-plugin/plugin.json Supports: install
- [Install instructions and skill index ↗](https://github.com/google-gemini/gemini-skills/blob/ee9da52195752ca83c2fca701d96bcb5a9c08d2d/README.md) Checked 2026-10-13 https://github.com/google-gemini/gemini-skills/blob/ee9da52195752ca83c2fca701d96bcb5a9c08d2d/README.md Supports: install, compatibility
- [Repository licence ↗](https://github.com/google-gemini/gemini-skills/blob/ee9da52195752ca83c2fca701d96bcb5a9c08d2d/LICENSE) Checked 2026-10-13 https://github.com/google-gemini/gemini-skills/blob/ee9da52195752ca83c2fca701d96bcb5a9c08d2d/LICENSE Supports: license
- [Pinned continuation review W41 ↗](https://raw.githubusercontent.com/google-gemini/gemini-skills/ee9da52195752ca83c2fca701d96bcb5a9c08d2d/skills/gemini-live-api-dev/SKILL.md) Checked 2026-10-15 https://raw.githubusercontent.com/google-gemini/gemini-skills/ee9da52195752ca83c2fca701d96bcb5a9c08d2d/skills/gemini-live-api-dev/SKILL.md Supports: review.findings, bestFor

KEEP COMPARING

## Other approaches to consider.

Related by category or shared topics. These are alternatives to inspect, not a measured quality order.

 [### Gemini API Development ↗ Gemini Interactions API patterns for Python and TypeScript, with a dated model list and deprecated-model rules.](/skills/google-gemini-gemini-api-dev/)[### Agent Platform GenAI Inference ↗ Calls Gemini and third-party OpenMaaS models on Google Cloud Agent Platform, asking for model, project and region instead of defaulting.](/skills/google-agent-platform-inference/)[### Amazon Bedrock ↗ Bedrock model invocation, Knowledge Bases, Guardrails and AgentCore guidance, with quota, prompt-caching and cost-tracking workflows.](/skills/aws-amazon-bedrock/)

 [AI Tools ↗](/tools/)[Skills ↗](/skills/)[Agents ↗](/agents/)[MCP Servers ↗](/mcp-servers/)[Workflows ↗](/workflows/)

## Continue your investigation

 - [Choose a toolkit](/tools/)
- [Inspect execution hosts](/agents/)
- [Check tool access](/mcp-servers/)
- [Check project dependencies](/check-repo/)
