Merge remote-tracking branch 'refs/remotes/origin/main'

providers/factory.py

@@ -68,7 +68,7 @@ def _build_single_stt_provider(
 ) -> BaseSTT:
     normalized = provider.lower().strip()
     if normalized in {"openai", "whisper", "openai_whisper"}:
-        # Whisper accepts an 8 kHz WAV container; keep the exact AudioSocket PCM, no resampling.
+        # Whisper accepts the configured WAV container; keep the exact AudioSocket PCM, no resampling.
         return OpenAISTT(input_sample_rate_hz=input_sample_rate_hz)
     if normalized in {"elevenlabs", "eleven_labs", "scribe"}:
         return ElevenLabsSTT(

providers/stt.py

@@ -1010,8 +1010,8 @@ class YandexSpeechKitSTT(BaseSTT):
         api_base: str | None = None,
         language: str | None = None,
         topic: str | None = None,
-        input_sample_rate_hz: int = 8000,
-        target_sample_rate_hz: int = 8000,
+        input_sample_rate_hz: int = 16000,
+        target_sample_rate_hz: int = 16000,
         timeout_seconds: float | None = None,
     ) -> None:
         self._api_key = str(api_key if api_key is not None else _yandex_api_key()).strip()

providers/stt_openai.py

@@ -79,7 +79,7 @@ class OpenAISTT(BaseSTT):
         api_key: str | None = None,
         model: str | None = None,
         base_url: str | None = None,
-        input_sample_rate_hz: int = 8000,
+        input_sample_rate_hz: int = 16000,
         prompt: str | None = None,
         language: str | None = None,
         timeout_seconds: float | None = None,

providers/tts.py

@@ -119,11 +119,15 @@ class ElevenLabsTTS(BaseTTS):
             str(output_format or os.getenv("ELEVENLABS_TTS_OUTPUT_FORMAT", "pcm_16000")).strip().lower()
             or "pcm_16000"
         )
+        requested_sample_rate_hz = _sample_rate_from_pcm_format(requested_output_format)
         self._target_sample_rate_hz = int(
             target_sample_rate_hz
             if target_sample_rate_hz is not None
-            else (_sample_rate_from_pcm_format(requested_output_format) or 16000)
+            else (requested_sample_rate_hz or 16000)
         )
+        if self._target_sample_rate_hz == 8000:
+            LOGGER.warning("ElevenLabs TTS target sample rate 8000Hz is disabled; using 16000Hz")
+            self._target_sample_rate_hz = 16000
         self._output_format = self._resolve_provider_output_format(
             requested_output_format=requested_output_format,
             target_sample_rate_hz=self._target_sample_rate_hz,
@@ -381,8 +385,7 @@ class ElevenLabsTTS(BaseTTS):
         normalized = requested_output_format.strip().lower() or "pcm_16000"
         if normalized == "pcm_8000":
             LOGGER.warning(
-                "ElevenLabs TTS does not provide reliable PCM16 8kHz streaming; requesting pcm_16000 "
-                "and resampling to %sHz locally",
+                "ElevenLabs TTS PCM16 8kHz output is disabled; requesting pcm_16000 for %sHz target",
                 target_sample_rate_hz,
             )
             return "pcm_16000"