Limit 1
The Kannada pack's colloquial register may need a client-specific override for a formal institution or a different regional variety.
ಕನ್ನಡ · measured India voice evidence
Kannada AI voice: runtime path, real pack lines and measured latency
Kannada resolves to Bharat Essential, Google Cloud Text-to-Speech voice kn-IN-Wavenet-A, the kn-IN dedicated pack and the cascade engine. Three fresh production-adapter runs measured a 900 ms median to the first non-empty 8 kHz PCM audio chunk on 15 Aug 2026.
Measured 15 Aug 2026 · three runs · 8 kHz PCM16 · no provider fallback
Which pack, engine and vendor voice serve Kannada?
The runtime trace starts with BaseBusiness.get_voice_pack, continues through voice_config.resolve_pack and ends at voice_config.resolve_engine. The generic integration returned none from the generic business integration, resolve_pack selected the kn-IN dedicated pack, and resolve_engine selected cascade. The TTS factory then selected Google Cloud Text-to-Speech voice kn-IN-Wavenet-A, displayed as Bhoomi, while speech recognition remained Sarvam saaras:v3.
| Runtime check | Resolved value | Meaning |
|---|---|---|
| get_voice_pack | None from the generic business integration | A client-specific pack would override the generic registry result. |
| resolve_pack | kn-IN dedicated pack | Fixed lines and language rules come from a locale-specific pack. |
| voice_config | Bharat Essential · Google Cloud Text-to-Speech · kn-IN-Wavenet-A | The grade write-through pins the production TTS provider and voice identifier. |
| resolve_engine | cascade | The cascade engine uses separate STT, reasoning and TTS stages. |
| speech recognition | Sarvam saaras:v3 | The primary recognition locale is kn-IN. |
The runtime table names the actual Kannada call path rather than a language-support logo: generic integration result, resolved pack, sold grade, provider voice, engine and speech-recognition provider. A client integration can replace the generic pack, but the benchmark fixture used no client override.
What first-audio latency did Kannada measure?
Kannada measured 870, 900, 938 ms across three isolated production-adapter requests. The median was 900 ms. Each run synthesized “ನಮಸ್ಕಾರ, ನಾನು ನಿಮಗೆ ಹೇಗೆ ಸಹಾಯ ಮಾಡಲಿ?” at 8 kHz PCM16, started the timer immediately before speak(text), stopped at the first non-empty audio chunk, verified the configured adapter had not failed over, and cancelled the iterator after that first chunk.
| Run | First audio | Audio format | Adapter result |
|---|---|---|---|
| Run 1 | 870 ms | 8 kHz PCM16 | Google Cloud Text-to-Speech; no fallback |
| Run 2 | 900 ms | 8 kHz PCM16 | Google Cloud Text-to-Speech; no fallback |
| Run 3 | 938 ms | 8 kHz PCM16 | Google Cloud Text-to-Speech; no fallback |
The latency table reports a median of 900 ms from three measurements on 15 Aug 2026. The number is a reproducible TTS-adapter read, not end-to-end call latency, a percentile, or an uptime promise. Caller speech, transcription, model generation, carrier delivery and handset playback remain outside the timer.
What register does the resolved Kannada pack use?
The dedicated Kannada pack uses conversational service Kannada with ರೀ and spoken forms such as ತಗೊಳ್ತೀನಿ. The register is deliberately closer to a phone conversation than formal written Kannada.
| Pack event | Exact runtime string | Register read |
|---|---|---|
| Filler while a tool runs | ಒಂದ್ ನಿಮಿಷ ರೀ | A conversational hold line with ರೀ. |
| Retry after unclear speech | ಸಾರಿ ರೀ, ಇನ್ನೊಂದ್ ಸಲ ಹೇಳ್ತೀರಾ? | Colloquial polite Kannada rather than a formal written retry. |
| Retract an unsupported action | ಸಾರಿ ರೀ, ಅದು ಇಲ್ಲಿಂದ ಕಳಿಸೋಕೆ ಆಗಲ್ಲ. ನಿಮ್ಮ ನಂಬರ್ ತಗೊಳ್ತೀನಿ, ನಮ್ಮವರು ಅದನ್ನ ನಿಮಗೆ ಕಳಿಸ್ತಾರೆ. | Spoken Kannada with a human follow-up path. |
The sample table copies three fixed strings from the kn-IN dedicated pack: the tool filler, unclear-speech retry and unsupported-action retraction. Kannada therefore has explicit runtime register evidence.
What are the honest Kannada limits?
Limit 2
The WaveNet adapter is non-streaming: Google returns the full clip before the first 200 ms PCM frame is yielded.
Limit 3
The three runs were tightly grouped at 870, 900 and 938 ms, but the median is not a production percentile or SLA.
Limit 4
The first-audio benchmark measures TTS adapter latency only; it excludes caller speech recognition, model reasoning, carrier transit and handset playback.
Which evidence supports the Kannada hub?
The fresh page-level pull did not expose a Kannada row in the selected output, so the Kannada hub consolidates the existing cost page without creating city or vertical combinations. The existing language article remains the workflow and cost depth page; the multilingual category index provides cross-language buying context; and the product page describes the broader scheduled-language surface.
Other measured Indian-language hubs
The related hubs keep one route per language and never multiply languages by verticals. Every page publishes its own runtime trace, three dated latency reads and fixed-pack evidence.
Kannada voice questions
The answers preserve the runtime resolution, measured median, exact run values and pack limitation as standalone passages that do not depend on the tables.
Kannada resolves to Bharat Essential, the Google Cloud Text-to-Speech kn-IN-Wavenet-A voice, Sarvam saaras:v3 recognition and the cascade engine. The generic integration returned no client-specific pack, so resolve_pack selected the kn-IN dedicated pack on 15 Aug 2026.
Kannada measured 900 ms median first-audio latency across three fresh 8 kHz PCM16 adapter runs: 870, 900, 938 ms. The timer started immediately before the production adapter's speak call and stopped at the first non-empty audio chunk on 15 Aug 2026.
Kannada uses the kn-IN dedicated pack. The fixed pack strings shown on this page are copied from the runtime object rather than written as marketing examples.