Add --no-vad option

README.md

@@ -34,6 +34,7 @@ whispering --language en --model tiny
 - ``--no-progress`` disables the progress message
 - ``-t`` sets temperatures to decode. You can set several like ``-t 0.0 -t 0.1 -t 0.5``, but too many temperatures exhaust decoding time
 - ``--debug`` outputs logs for debug
+- ``--no-vad`` disables VAD (Voice Activity Detection). This forces whisper to analyze non-voice activity sound period
 
 ### Parse interval
 

whispering/cli.py

@@ -155,6 +155,10 @@ def get_opts() -> argparse.Namespace:
         "--no-progress",
         action="store_true",
     )
+    parser.add_argument(
+        "--no-vad",
+        action="store_true",
+    )
     opts = parser.parse_args()
 
     if opts.beam_size <= 0:
@@ -187,6 +191,7 @@ def get_context(*, opts) -> Context:
         beam_size=opts.beam_size,
         temperatures=opts.temperature,
         allow_padding=opts.allow_padding,
+        vad=not opts.no_vad,
     )
     logger.debug(f"Context: {ctx}")
     return ctx

whispering/schema.py

@@ -27,6 +27,7 @@ class Context(BaseModel, arbitrary_types_allowed=True):
     timestamp: float = 0.0
     buffer_tokens: List[torch.Tensor] = []
     buffer_mel: Optional[torch.Tensor] = None
+    vad: bool = True
 
     temperatures: List[float]
     allow_padding: bool = False

whispering/transcriber.py

@@ -233,6 +233,8 @@ class WhisperStreamingTranscriber:
         ctx: Context,
     ) -> Iterator[ParsedChunk]:
         logger.debug(f"{len(audio)}")
+
+        if not ctx.vad:
             x = [
                 v
                 for v in self.vad(