Pipecat 1.16 adds PerceptionReader for Google Decision 1 audio
The notable shift is that paralinguistic signals like sarcasm and prosody become a first-class input for voice agents, something most ASR-plus-LLM pipelines discard entirely. Teams building customer-facing voice bots in Pipecat can test the capability immediately via the live demo and the 1.16 docs. Compared with conventional transcript-only agent stacks, this enables tone-aware routing and response strategies, useful for enterprise support teams evaluating Decision 1 against OpenAI realtime audio options.
