Speech to Text accepts vad_threshold to tune the voice-activity gate that skips non-speech audio.

The control is the gate, giving builders a way to trade silence filtering for noisier speech capture.

It sits in streaming and batch Speech to Text, where narrowband telephony needs quieter or noisier speech handled explicitly.

• Aug 05, 2026OpenAI removed model selection copy naming GPT-5.6 Sol, Terra, and Luna.
• Aug 05, 2026Anthropic removed model card copy for Claude Fable 5 and Claude Mythos 5.
• Aug 05, 2026Google pricing page removed Gemini model navigation and Interactions API availability note

This is the world we build in. It moves this fast. We move with it.