Add more Python examples for SenseVoice (#1179)

python-api-examples/generate-subtitles.py

@@ -12,12 +12,12 @@ Supported file formats are those supported by ffmpeg; for instance,
 Note that you need a non-streaming model for this script.
 
 Please visit
-https://github.com/snakers4/silero-vad/raw/master/src/silero_vad/data/silero_vad.onnx
+https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/silero_vad.onnx
 to download silero_vad.onnx
 
 For instance,
 
-wget https://github.com/snakers4/silero-vad/raw/master/src/silero_vad/data/silero_vad.onnx
+wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/silero_vad.onnx
 
 (1) For paraformer
 
@@ -58,7 +58,17 @@ wget https://github.com/snakers4/silero-vad/raw/master/src/silero_vad/data/siler
   --num-threads=2 \
   /path/to/test.mp4
 
-(4) For WeNet CTC models
+(4) For SenseVoice CTC models
+
+./python-api-examples/generate-subtitles.py  \
+  --silero-vad-model=/path/to/silero_vad.onnx \
+  --sense-voice=./sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17/model.onnx \
+  --tokens=./sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17/tokens.txt \
+  --num-threads=2 \
+  /path/to/test.mp4
+
+
+(5) For WeNet CTC models
 
 ./python-api-examples/generate-subtitles.py  \
   --silero-vad-model=/path/to/silero_vad.onnx \
@@ -130,6 +140,13 @@ def get_args():
         help="Path to the model.onnx from Paraformer",
     )
 
+    parser.add_argument(
+        "--sense-voice",
+        default="",
+        type=str,
+        help="Path to the model.onnx from SenseVoice",
+    )
+
     parser.add_argument(
         "--wenet-ctc",
         default="",
@@ -242,6 +259,7 @@ def assert_file_exists(filename: str):
 def create_recognizer(args) -> sherpa_onnx.OfflineRecognizer:
     if args.encoder:
         assert len(args.paraformer) == 0, args.paraformer
+        assert len(args.sense_voice) == 0, args.sense_voice
         assert len(args.wenet_ctc) == 0, args.wenet_ctc
         assert len(args.whisper_encoder) == 0, args.whisper_encoder
         assert len(args.whisper_decoder) == 0, args.whisper_decoder
@@ -262,6 +280,7 @@ def create_recognizer(args) -> sherpa_onnx.OfflineRecognizer:
             debug=args.debug,
         )
     elif args.paraformer:
+        assert len(args.sense_voice) == 0, args.sense_voice
         assert len(args.wenet_ctc) == 0, args.wenet_ctc
         assert len(args.whisper_encoder) == 0, args.whisper_encoder
         assert len(args.whisper_decoder) == 0, args.whisper_decoder
@@ -277,6 +296,19 @@ def create_recognizer(args) -> sherpa_onnx.OfflineRecognizer:
             decoding_method=args.decoding_method,
             debug=args.debug,
         )
+    elif args.sense_voice:
+        assert len(args.wenet_ctc) == 0, args.wenet_ctc
+        assert len(args.whisper_encoder) == 0, args.whisper_encoder
+        assert len(args.whisper_decoder) == 0, args.whisper_decoder
+
+        assert_file_exists(args.sense_voice)
+        recognizer = sherpa_onnx.OfflineRecognizer.from_sense_voice(
+            model=args.sense_voice,
+            tokens=args.tokens,
+            num_threads=args.num_threads,
+            use_itn=True,
+            debug=args.debug,
+        )
     elif args.wenet_ctc:
         assert len(args.whisper_encoder) == 0, args.whisper_encoder
         assert len(args.whisper_decoder) == 0, args.whisper_decoder
@@ -406,6 +438,9 @@ def main():
             vad.accept_waveform(buffer[:window_size])
             buffer = buffer[window_size:]
 
+        if is_silence:
+            vad.flush()
+
         streams = []
         segments = []
         while not vad.empty():