A good fit for
- → Testing multimodal prompts
- → Prototyping voice and speech generation
- → Inspecting model behavior before backend integration
Know the limits
- → Account, region and model access can differ.
- → Flash text output is not the same as Live audio output.
- → Preview behavior and production quotas should be tested separately.
Your first useful workflow
Pick the modality
Decide whether you need text generation, live dialogue or scripted speech.
Use representative inputs
Include a difficult document, an ambiguous request and a failure case.
Move secrets server-side
Keep the API key out of your browser application and add request limits.
The practical take
Flash, Live and TTS share a family name, not one interchangeable request format.
Sources & evidence
Capabilities above are summarized from these official documents. Workflow suggestions are editorial guidance, not a claim of measured product superiority.