Tom Osman
←[ RETURN_TO_INVENTORY ]

Google AI Studio

// February 20, 2025

Google AI Studio homepage
Live preview captured from aistudio.google.com

Google calls AI Studio "the fastest place to start building with the Gemini API," and the latest release doubled down on that promise with native code generation, an upgraded media lab, and richer tooling. (Developers blog, May 21, 2025)

Why I keep a window open

  • Build tab = code IDE. Gemini 2.5 Pro now powers AI Studio's "Build" tab, which can generate full web apps from a text/image/video prompt, iterate over chat with diffs and checkpoints, and deploy straight to Cloud Run with a click. Since AI Studio proxies Gemini calls through a placeholder key, people viewing the shared app use their own free quota instead of mine.
  • Generate Media hub. Imagen, Veo, Gemini image generation, speech models, and even Lyria PromptDJ now sit under a single "Generate Media" page, so I can test video, audio, and music models without juggling tabs.
  • Live API + TTS upgrades. Gemini 2.5 Flash's native audio dialog adds over 30 voices, "proactive audio" to ignore background chatter, and new text-to-speech options for crafting single or multi-speaker output (perfect for conversational interfaces).
  • Agent tooling. There's now native Model Context Protocol (MCP) support in the Gen AI SDK plus demo apps (like a Google Maps MCP server) and an experimental URL Context tool that lets Gemini retrieve content from links for fact-checking and summaries.

Plays I run with it

  1. Spec-to-app demos. Drop a product brief into Build, let Gemini scaffold the frontend/backend, and use the chat diffing to steer revisions before deploying to Cloud Run.
  2. Multimodal proofs. Hop into Generate Media to pull a Veo shot list, Imagen assets, then wire Gemini native image generation into the same prompt to see if the creative direction holds.
  3. Voice-first flows. Prototype hotline or kiosk dialogs with Gemini 2.5 Flash in the Stream tab, tweak the TTS settings, and record the results for stakeholders before writing any client code.

Getting in

  • AI Studio runs in the browser at aistudio.google.com; sign in with your Google account to access the Build, Generate Media, Stream, and Live tabs.
  • Generated code uses a placeholder API key until you swap in your own credentials, so always review before you deploy externally.
  • Keep the announcement post handy; it's the canonical reference for what models/features have landed in the current preview.
Open Google AI Studio