{"language":"Hindi","locale":"hi-IN","measuredAt":"2026-08-15","method":"Three isolated production make_tts adapter instances at 8 kHz PCM16; await Azure built-in connection/model warm-up; start perf_counter immediately before speak(text); stop at first non-empty audio chunk; cancel after first chunk; report median; exclude STT, model, carrier and playback.","runtime":{"grade":"Bharat Essential","provider":"Google Cloud Text-to-Speech","vendorVoice":"hi-IN-Wavenet-A","displayVoice":"Kavya","sttProvider":"Sarvam saaras:v3","engine":"cascade","getVoicePackResult":"None from the generic business integration","resolvedPack":"hi-IN dedicated pack"},"benchmarkText":"नमस्ते, मैं आपकी कैसे मदद कर सकती हूँ?","latencyRunsMs":[788,822,946],"medianFirstAudioMs":822,"packSamples":[{"purpose":"Filler while a tool runs","text":"एक पल","registerNote":"A compact neutral Hindi hold line."},{"purpose":"Retry after unclear speech","text":"माफ़ कीजिए, एक बार फिर बताइएगा?","registerNote":"Polite service Hindi rather than casual तू/तुम register."},{"purpose":"Retract an unsupported action","text":"माफ़ कीजिए, वो मैं यहाँ से भेज नहीं सकती। आपका नंबर ले लेती हूँ, कोई आपको भेज देगा।","registerNote":"Feminine agent form with a human fallback."}],"honestLimits":["The WaveNet adapter is non-streaming: Google returns the full clip before the first 200 ms PCM frame is yielded.","The benchmark uses one short Hindi sentence and three isolated requests; 822 ms is a reproducible median, not a production percentile or SLA.","A dedicated Hindi pack does not prove every business name, English product name or regional accent will be pronounced correctly.","The first-audio benchmark measures TTS adapter latency only; it excludes caller speech recognition, model reasoning, carrier transit and handset playback."]}